From 2488f6b4c0ef0a72758cf58e5c9b419fefa8f10e Mon Sep 17 00:00:00 2001 From: Nathaniel Cook Date: Thu, 9 Jun 2016 12:55:40 -0600 Subject: [PATCH] Squashed 'vendor/github.com/BurntSushi/toml/' content from commit 2ceedfe git-subtree-dir: vendor/github.com/BurntSushi/toml git-subtree-split: 2ceedfee35ad3848e49308ab0c9a4f640cfb5fb2 --- .gitignore | 5 + .travis.yml | 12 + COMPATIBLE | 3 + COPYING | 14 + Makefile | 19 + README.md | 220 ++++++++++ _examples/example.go | 59 +++ _examples/example.toml | 35 ++ _examples/hard.toml | 22 + _examples/implicit.toml | 4 + _examples/invalid-apples.toml | 6 + _examples/invalid.toml | 35 ++ _examples/readme1.toml | 5 + _examples/readme2.toml | 1 + cmd/toml-test-decoder/COPYING | 14 + cmd/toml-test-decoder/README.md | 14 + cmd/toml-test-decoder/main.go | 90 ++++ cmd/toml-test-encoder/COPYING | 14 + cmd/toml-test-encoder/README.md | 14 + cmd/toml-test-encoder/main.go | 131 ++++++ cmd/tomlv/COPYING | 14 + cmd/tomlv/README.md | 22 + cmd/tomlv/main.go | 61 +++ decode.go | 472 ++++++++++++++++++++ decode_meta.go | 99 +++++ decode_test.go | 540 +++++++++++++++++++++++ doc.go | 27 ++ encode.go | 515 ++++++++++++++++++++++ encode_test.go | 506 ++++++++++++++++++++++ encoding_types.go | 19 + encoding_types_1.1.go | 18 + lex.go | 734 ++++++++++++++++++++++++++++++++ parse.go | 417 ++++++++++++++++++ session.vim | 1 + type_check.go | 85 ++++ type_fields.go | 241 +++++++++++ 36 files changed, 4488 insertions(+) create mode 100644 .gitignore create mode 100644 .travis.yml create mode 100644 COMPATIBLE create mode 100644 COPYING create mode 100644 Makefile create mode 100644 README.md create mode 100644 _examples/example.go create mode 100644 _examples/example.toml create mode 100644 _examples/hard.toml create mode 100644 _examples/implicit.toml create mode 100644 _examples/invalid-apples.toml create mode 100644 _examples/invalid.toml create mode 100644 _examples/readme1.toml create mode 100644 _examples/readme2.toml create mode 100644 cmd/toml-test-decoder/COPYING create mode 100644 cmd/toml-test-decoder/README.md create mode 100644 cmd/toml-test-decoder/main.go create mode 100644 cmd/toml-test-encoder/COPYING create mode 100644 cmd/toml-test-encoder/README.md create mode 100644 cmd/toml-test-encoder/main.go create mode 100644 cmd/tomlv/COPYING create mode 100644 cmd/tomlv/README.md create mode 100644 cmd/tomlv/main.go create mode 100644 decode.go create mode 100644 decode_meta.go create mode 100644 decode_test.go create mode 100644 doc.go create mode 100644 encode.go create mode 100644 encode_test.go create mode 100644 encoding_types.go create mode 100644 encoding_types_1.1.go create mode 100644 lex.go create mode 100644 parse.go create mode 100644 session.vim create mode 100644 type_check.go create mode 100644 type_fields.go diff --git a/.gitignore b/.gitignore new file mode 100644 index 000000000..0cd380037 --- /dev/null +++ b/.gitignore @@ -0,0 +1,5 @@ +TAGS +tags +.*.swp +tomlcheck/tomlcheck +toml.test diff --git a/.travis.yml b/.travis.yml new file mode 100644 index 000000000..43caf6d02 --- /dev/null +++ b/.travis.yml @@ -0,0 +1,12 @@ +language: go +go: + - 1.1 + - 1.2 + - tip +install: + - go install ./... + - go get github.com/BurntSushi/toml-test +script: + - export PATH="$PATH:$HOME/gopath/bin" + - make test + diff --git a/COMPATIBLE b/COMPATIBLE new file mode 100644 index 000000000..21e0938ca --- /dev/null +++ b/COMPATIBLE @@ -0,0 +1,3 @@ +Compatible with TOML version +[v0.2.0](https://github.com/mojombo/toml/blob/master/versions/toml-v0.2.0.md) + diff --git a/COPYING b/COPYING new file mode 100644 index 000000000..5a8e33254 --- /dev/null +++ b/COPYING @@ -0,0 +1,14 @@ + DO WHAT THE FUCK YOU WANT TO PUBLIC LICENSE + Version 2, December 2004 + + Copyright (C) 2004 Sam Hocevar + + Everyone is permitted to copy and distribute verbatim or modified + copies of this license document, and changing it is allowed as long + as the name is changed. + + DO WHAT THE FUCK YOU WANT TO PUBLIC LICENSE + TERMS AND CONDITIONS FOR COPYING, DISTRIBUTION AND MODIFICATION + + 0. You just DO WHAT THE FUCK YOU WANT TO. + diff --git a/Makefile b/Makefile new file mode 100644 index 000000000..3600848d3 --- /dev/null +++ b/Makefile @@ -0,0 +1,19 @@ +install: + go install ./... + +test: install + go test -v + toml-test toml-test-decoder + toml-test -encoder toml-test-encoder + +fmt: + gofmt -w *.go */*.go + colcheck *.go */*.go + +tags: + find ./ -name '*.go' -print0 | xargs -0 gotags > TAGS + +push: + git push origin master + git push github master + diff --git a/README.md b/README.md new file mode 100644 index 000000000..380bb36bb --- /dev/null +++ b/README.md @@ -0,0 +1,220 @@ +## TOML parser and encoder for Go with reflection + +TOML stands for Tom's Obvious, Minimal Language. This Go package provides a +reflection interface similar to Go's standard library `json` and `xml` +packages. This package also supports the `encoding.TextUnmarshaler` and +`encoding.TextMarshaler` interfaces so that you can define custom data +representations. (There is an example of this below.) + +Spec: https://github.com/mojombo/toml + +Compatible with TOML version +[v0.2.0](https://github.com/mojombo/toml/blob/master/versions/toml-v0.2.0.md) + +Documentation: http://godoc.org/github.com/BurntSushi/toml + +Installation: + +```bash +go get github.com/BurntSushi/toml +``` + +Try the toml validator: + +```bash +go get github.com/BurntSushi/toml/cmd/tomlv +tomlv some-toml-file.toml +``` + +[![Build status](https://api.travis-ci.org/BurntSushi/toml.png)](https://travis-ci.org/BurntSushi/toml) + + +### Testing + +This package passes all tests in +[toml-test](https://github.com/BurntSushi/toml-test) for both the decoder +and the encoder. + +### Examples + +This package works similarly to how the Go standard library handles `XML` +and `JSON`. Namely, data is loaded into Go values via reflection. + +For the simplest example, consider some TOML file as just a list of keys +and values: + +```toml +Age = 25 +Cats = [ "Cauchy", "Plato" ] +Pi = 3.14 +Perfection = [ 6, 28, 496, 8128 ] +DOB = 1987-07-05T05:45:00Z +``` + +Which could be defined in Go as: + +```go +type Config struct { + Age int + Cats []string + Pi float64 + Perfection []int + DOB time.Time // requires `import time` +} +``` + +And then decoded with: + +```go +var conf Config +if _, err := toml.Decode(tomlData, &conf); err != nil { + // handle error +} +``` + +You can also use struct tags if your struct field name doesn't map to a TOML +key value directly: + +```toml +some_key_NAME = "wat" +``` + +```go +type TOML struct { + ObscureKey string `toml:"some_key_NAME"` +} +``` + +### Using the `encoding.TextUnmarshaler` interface + +Here's an example that automatically parses duration strings into +`time.Duration` values: + +```toml +[[song]] +name = "Thunder Road" +duration = "4m49s" + +[[song]] +name = "Stairway to Heaven" +duration = "8m03s" +``` + +Which can be decoded with: + +```go +type song struct { + Name string + Duration duration +} +type songs struct { + Song []song +} +var favorites songs +if _, err := Decode(blob, &favorites); err != nil { + log.Fatal(err) +} + +for _, s := range favorites.Song { + fmt.Printf("%s (%s)\n", s.Name, s.Duration) +} +``` + +And you'll also need a `duration` type that satisfies the +`encoding.TextUnmarshaler` interface: + +```go +type duration struct { + time.Duration +} + +func (d *duration) UnmarshalText(text []byte) error { + var err error + d.Duration, err = time.ParseDuration(string(text)) + return err +} +``` + +### More complex usage + +Here's an example of how to load the example from the official spec page: + +```toml +# This is a TOML document. Boom. + +title = "TOML Example" + +[owner] +name = "Tom Preston-Werner" +organization = "GitHub" +bio = "GitHub Cofounder & CEO\nLikes tater tots and beer." +dob = 1979-05-27T07:32:00Z # First class dates? Why not? + +[database] +server = "192.168.1.1" +ports = [ 8001, 8001, 8002 ] +connection_max = 5000 +enabled = true + +[servers] + + # You can indent as you please. Tabs or spaces. TOML don't care. + [servers.alpha] + ip = "10.0.0.1" + dc = "eqdc10" + + [servers.beta] + ip = "10.0.0.2" + dc = "eqdc10" + +[clients] +data = [ ["gamma", "delta"], [1, 2] ] # just an update to make sure parsers support it + +# Line breaks are OK when inside arrays +hosts = [ + "alpha", + "omega" +] +``` + +And the corresponding Go types are: + +```go +type tomlConfig struct { + Title string + Owner ownerInfo + DB database `toml:"database"` + Servers map[string]server + Clients clients +} + +type ownerInfo struct { + Name string + Org string `toml:"organization"` + Bio string + DOB time.Time +} + +type database struct { + Server string + Ports []int + ConnMax int `toml:"connection_max"` + Enabled bool +} + +type server struct { + IP string + DC string +} + +type clients struct { + Data [][]interface{} + Hosts []string +} +``` + +Note that a case insensitive match will be tried if an exact match can't be +found. + +A working example of the above can be found in `_examples/example.{go,toml}`. + diff --git a/_examples/example.go b/_examples/example.go new file mode 100644 index 000000000..c81d25a52 --- /dev/null +++ b/_examples/example.go @@ -0,0 +1,59 @@ +package main + +import ( + "fmt" + "time" + + "github.com/BurntSushi/toml" +) + +type tomlConfig struct { + Title string + Owner ownerInfo + DB database `toml:"database"` + Servers map[string]server + Clients clients +} + +type ownerInfo struct { + Name string + Org string `toml:"organization"` + Bio string + DOB time.Time +} + +type database struct { + Server string + Ports []int + ConnMax int `toml:"connection_max"` + Enabled bool +} + +type server struct { + IP string + DC string +} + +type clients struct { + Data [][]interface{} + Hosts []string +} + +func main() { + var config tomlConfig + if _, err := toml.DecodeFile("example.toml", &config); err != nil { + fmt.Println(err) + return + } + + fmt.Printf("Title: %s\n", config.Title) + fmt.Printf("Owner: %s (%s, %s), Born: %s\n", + config.Owner.Name, config.Owner.Org, config.Owner.Bio, config.Owner.DOB) + fmt.Printf("Database: %s %v (Max conn. %d), Enabled? %v\n", + config.DB.Server, config.DB.Ports, config.DB.ConnMax, config.DB.Enabled) + for serverName, server := range config.Servers { + fmt.Printf("Server: %s (%s, %s)\n", serverName, server.IP, server.DC) + } + fmt.Printf("Client data: %v\n", config.Clients.Data) + fmt.Printf("Client hosts: %v\n", config.Clients.Hosts) +} diff --git a/_examples/example.toml b/_examples/example.toml new file mode 100644 index 000000000..32c7a4faa --- /dev/null +++ b/_examples/example.toml @@ -0,0 +1,35 @@ +# This is a TOML document. Boom. + +title = "TOML Example" + +[owner] +name = "Tom Preston-Werner" +organization = "GitHub" +bio = "GitHub Cofounder & CEO\nLikes tater tots and beer." +dob = 1979-05-27T07:32:00Z # First class dates? Why not? + +[database] +server = "192.168.1.1" +ports = [ 8001, 8001, 8002 ] +connection_max = 5000 +enabled = true + +[servers] + + # You can indent as you please. Tabs or spaces. TOML don't care. + [servers.alpha] + ip = "10.0.0.1" + dc = "eqdc10" + + [servers.beta] + ip = "10.0.0.2" + dc = "eqdc10" + +[clients] +data = [ ["gamma", "delta"], [1, 2] ] # just an update to make sure parsers support it + +# Line breaks are OK when inside arrays +hosts = [ + "alpha", + "omega" +] diff --git a/_examples/hard.toml b/_examples/hard.toml new file mode 100644 index 000000000..26145d2b4 --- /dev/null +++ b/_examples/hard.toml @@ -0,0 +1,22 @@ +# Test file for TOML +# Only this one tries to emulate a TOML file written by a user of the kind of parser writers probably hate +# This part you'll really hate + +[the] +test_string = "You'll hate me after this - #" # " Annoying, isn't it? + + [the.hard] + test_array = [ "] ", " # "] # ] There you go, parse this! + test_array2 = [ "Test #11 ]proved that", "Experiment #9 was a success" ] + # You didn't think it'd as easy as chucking out the last #, did you? + another_test_string = " Same thing, but with a string #" + harder_test_string = " And when \"'s are in the string, along with # \"" # "and comments are there too" + # Things will get harder + + [the.hard.bit#] + what? = "You don't think some user won't do that?" + multi_line_array = [ + "]", + # ] Oh yes I did + ] + diff --git a/_examples/implicit.toml b/_examples/implicit.toml new file mode 100644 index 000000000..1dea5ceb4 --- /dev/null +++ b/_examples/implicit.toml @@ -0,0 +1,4 @@ +# [x] you +# [x.y] don't +# [x.y.z] need these +[x.y.z.w] # for this to work diff --git a/_examples/invalid-apples.toml b/_examples/invalid-apples.toml new file mode 100644 index 000000000..74e9e337e --- /dev/null +++ b/_examples/invalid-apples.toml @@ -0,0 +1,6 @@ +# DO NOT WANT +[fruit] +type = "apple" + +[fruit.type] +apple = "yes" diff --git a/_examples/invalid.toml b/_examples/invalid.toml new file mode 100644 index 000000000..beb1dba54 --- /dev/null +++ b/_examples/invalid.toml @@ -0,0 +1,35 @@ +# This is an INVALID TOML document. Boom. +# Can you spot the error without help? + +title = "TOML Example" + +[owner] +name = "Tom Preston-Werner" +organization = "GitHub" +bio = "GitHub Cofounder & CEO\nLikes tater tots and beer." +dob = 1979-05-27T7:32:00Z # First class dates? Why not? + +[database] +server = "192.168.1.1" +ports = [ 8001, 8001, 8002 ] +connection_max = 5000 +enabled = true + +[servers] + # You can indent as you please. Tabs or spaces. TOML don't care. + [servers.alpha] + ip = "10.0.0.1" + dc = "eqdc10" + + [servers.beta] + ip = "10.0.0.2" + dc = "eqdc10" + +[clients] +data = [ ["gamma", "delta"], [1, 2] ] # just an update to make sure parsers support it + +# Line breaks are OK when inside arrays +hosts = [ + "alpha", + "omega" +] diff --git a/_examples/readme1.toml b/_examples/readme1.toml new file mode 100644 index 000000000..3e1261d4c --- /dev/null +++ b/_examples/readme1.toml @@ -0,0 +1,5 @@ +Age = 25 +Cats = [ "Cauchy", "Plato" ] +Pi = 3.14 +Perfection = [ 6, 28, 496, 8128 ] +DOB = 1987-07-05T05:45:00Z diff --git a/_examples/readme2.toml b/_examples/readme2.toml new file mode 100644 index 000000000..b51cd9340 --- /dev/null +++ b/_examples/readme2.toml @@ -0,0 +1 @@ +some_key_NAME = "wat" diff --git a/cmd/toml-test-decoder/COPYING b/cmd/toml-test-decoder/COPYING new file mode 100644 index 000000000..5a8e33254 --- /dev/null +++ b/cmd/toml-test-decoder/COPYING @@ -0,0 +1,14 @@ + DO WHAT THE FUCK YOU WANT TO PUBLIC LICENSE + Version 2, December 2004 + + Copyright (C) 2004 Sam Hocevar + + Everyone is permitted to copy and distribute verbatim or modified + copies of this license document, and changing it is allowed as long + as the name is changed. + + DO WHAT THE FUCK YOU WANT TO PUBLIC LICENSE + TERMS AND CONDITIONS FOR COPYING, DISTRIBUTION AND MODIFICATION + + 0. You just DO WHAT THE FUCK YOU WANT TO. + diff --git a/cmd/toml-test-decoder/README.md b/cmd/toml-test-decoder/README.md new file mode 100644 index 000000000..24421eb70 --- /dev/null +++ b/cmd/toml-test-decoder/README.md @@ -0,0 +1,14 @@ +# Implements the TOML test suite interface + +This is an implementation of the interface expected by +[toml-test](https://github.com/BurntSushi/toml-test) for my +[toml parser written in Go](https://github.com/BurntSushi/toml). +In particular, it maps TOML data on `stdin` to a JSON format on `stdout`. + + +Compatible with TOML version +[v0.2.0](https://github.com/mojombo/toml/blob/master/versions/toml-v0.2.0.md) + +Compatible with `toml-test` version +[v0.2.0](https://github.com/BurntSushi/toml-test/tree/v0.2.0) + diff --git a/cmd/toml-test-decoder/main.go b/cmd/toml-test-decoder/main.go new file mode 100644 index 000000000..14e755700 --- /dev/null +++ b/cmd/toml-test-decoder/main.go @@ -0,0 +1,90 @@ +// Command toml-test-decoder satisfies the toml-test interface for testing +// TOML decoders. Namely, it accepts TOML on stdin and outputs JSON on stdout. +package main + +import ( + "encoding/json" + "flag" + "fmt" + "log" + "os" + "path" + "time" + + "github.com/BurntSushi/toml" +) + +func init() { + log.SetFlags(0) + + flag.Usage = usage + flag.Parse() +} + +func usage() { + log.Printf("Usage: %s < toml-file\n", path.Base(os.Args[0])) + flag.PrintDefaults() + + os.Exit(1) +} + +func main() { + if flag.NArg() != 0 { + flag.Usage() + } + + var tmp interface{} + if _, err := toml.DecodeReader(os.Stdin, &tmp); err != nil { + log.Fatalf("Error decoding TOML: %s", err) + } + + typedTmp := translate(tmp) + if err := json.NewEncoder(os.Stdout).Encode(typedTmp); err != nil { + log.Fatalf("Error encoding JSON: %s", err) + } +} + +func translate(tomlData interface{}) interface{} { + switch orig := tomlData.(type) { + case map[string]interface{}: + typed := make(map[string]interface{}, len(orig)) + for k, v := range orig { + typed[k] = translate(v) + } + return typed + case []map[string]interface{}: + typed := make([]map[string]interface{}, len(orig)) + for i, v := range orig { + typed[i] = translate(v).(map[string]interface{}) + } + return typed + case []interface{}: + typed := make([]interface{}, len(orig)) + for i, v := range orig { + typed[i] = translate(v) + } + + // We don't really need to tag arrays, but let's be future proof. + // (If TOML ever supports tuples, we'll need this.) + return tag("array", typed) + case time.Time: + return tag("datetime", orig.Format("2006-01-02T15:04:05Z")) + case bool: + return tag("bool", fmt.Sprintf("%v", orig)) + case int64: + return tag("integer", fmt.Sprintf("%d", orig)) + case float64: + return tag("float", fmt.Sprintf("%v", orig)) + case string: + return tag("string", orig) + } + + panic(fmt.Sprintf("Unknown type: %T", tomlData)) +} + +func tag(typeName string, data interface{}) map[string]interface{} { + return map[string]interface{}{ + "type": typeName, + "value": data, + } +} diff --git a/cmd/toml-test-encoder/COPYING b/cmd/toml-test-encoder/COPYING new file mode 100644 index 000000000..5a8e33254 --- /dev/null +++ b/cmd/toml-test-encoder/COPYING @@ -0,0 +1,14 @@ + DO WHAT THE FUCK YOU WANT TO PUBLIC LICENSE + Version 2, December 2004 + + Copyright (C) 2004 Sam Hocevar + + Everyone is permitted to copy and distribute verbatim or modified + copies of this license document, and changing it is allowed as long + as the name is changed. + + DO WHAT THE FUCK YOU WANT TO PUBLIC LICENSE + TERMS AND CONDITIONS FOR COPYING, DISTRIBUTION AND MODIFICATION + + 0. You just DO WHAT THE FUCK YOU WANT TO. + diff --git a/cmd/toml-test-encoder/README.md b/cmd/toml-test-encoder/README.md new file mode 100644 index 000000000..45a603f29 --- /dev/null +++ b/cmd/toml-test-encoder/README.md @@ -0,0 +1,14 @@ +# Implements the TOML test suite interface for TOML encoders + +This is an implementation of the interface expected by +[toml-test](https://github.com/BurntSushi/toml-test) for the +[TOML encoder](https://github.com/BurntSushi/toml). +In particular, it maps JSON data on `stdin` to a TOML format on `stdout`. + + +Compatible with TOML version +[v0.2.0](https://github.com/mojombo/toml/blob/master/versions/toml-v0.2.0.md) + +Compatible with `toml-test` version +[v0.2.0](https://github.com/BurntSushi/toml-test/tree/v0.2.0) + diff --git a/cmd/toml-test-encoder/main.go b/cmd/toml-test-encoder/main.go new file mode 100644 index 000000000..092cc6844 --- /dev/null +++ b/cmd/toml-test-encoder/main.go @@ -0,0 +1,131 @@ +// Command toml-test-encoder satisfies the toml-test interface for testing +// TOML encoders. Namely, it accepts JSON on stdin and outputs TOML on stdout. +package main + +import ( + "encoding/json" + "flag" + "log" + "os" + "path" + "strconv" + "time" + + "github.com/BurntSushi/toml" +) + +func init() { + log.SetFlags(0) + + flag.Usage = usage + flag.Parse() +} + +func usage() { + log.Printf("Usage: %s < json-file\n", path.Base(os.Args[0])) + flag.PrintDefaults() + + os.Exit(1) +} + +func main() { + if flag.NArg() != 0 { + flag.Usage() + } + + var tmp interface{} + if err := json.NewDecoder(os.Stdin).Decode(&tmp); err != nil { + log.Fatalf("Error decoding JSON: %s", err) + } + + tomlData := translate(tmp) + if err := toml.NewEncoder(os.Stdout).Encode(tomlData); err != nil { + log.Fatalf("Error encoding TOML: %s", err) + } +} + +func translate(typedJson interface{}) interface{} { + switch v := typedJson.(type) { + case map[string]interface{}: + if len(v) == 2 && in("type", v) && in("value", v) { + return untag(v) + } + m := make(map[string]interface{}, len(v)) + for k, v2 := range v { + m[k] = translate(v2) + } + return m + case []interface{}: + tabArray := make([]map[string]interface{}, len(v)) + for i := range v { + if m, ok := translate(v[i]).(map[string]interface{}); ok { + tabArray[i] = m + } else { + log.Fatalf("JSON arrays may only contain objects. This " + + "corresponds to only tables being allowed in " + + "TOML table arrays.") + } + } + return tabArray + } + log.Fatalf("Unrecognized JSON format '%T'.", typedJson) + panic("unreachable") +} + +func untag(typed map[string]interface{}) interface{} { + t := typed["type"].(string) + v := typed["value"] + switch t { + case "string": + return v.(string) + case "integer": + v := v.(string) + n, err := strconv.Atoi(v) + if err != nil { + log.Fatalf("Could not parse '%s' as integer: %s", v, err) + } + return n + case "float": + v := v.(string) + f, err := strconv.ParseFloat(v, 64) + if err != nil { + log.Fatalf("Could not parse '%s' as float64: %s", v, err) + } + return f + case "datetime": + v := v.(string) + t, err := time.Parse("2006-01-02T15:04:05Z", v) + if err != nil { + log.Fatalf("Could not parse '%s' as a datetime: %s", v, err) + } + return t + case "bool": + v := v.(string) + switch v { + case "true": + return true + case "false": + return false + } + log.Fatalf("Could not parse '%s' as a boolean.", v) + case "array": + v := v.([]interface{}) + array := make([]interface{}, len(v)) + for i := range v { + if m, ok := v[i].(map[string]interface{}); ok { + array[i] = untag(m) + } else { + log.Fatalf("Arrays may only contain other arrays or "+ + "primitive values, but found a '%T'.", m) + } + } + return array + } + log.Fatalf("Unrecognized tag type '%s'.", t) + panic("unreachable") +} + +func in(key string, m map[string]interface{}) bool { + _, ok := m[key] + return ok +} diff --git a/cmd/tomlv/COPYING b/cmd/tomlv/COPYING new file mode 100644 index 000000000..5a8e33254 --- /dev/null +++ b/cmd/tomlv/COPYING @@ -0,0 +1,14 @@ + DO WHAT THE FUCK YOU WANT TO PUBLIC LICENSE + Version 2, December 2004 + + Copyright (C) 2004 Sam Hocevar + + Everyone is permitted to copy and distribute verbatim or modified + copies of this license document, and changing it is allowed as long + as the name is changed. + + DO WHAT THE FUCK YOU WANT TO PUBLIC LICENSE + TERMS AND CONDITIONS FOR COPYING, DISTRIBUTION AND MODIFICATION + + 0. You just DO WHAT THE FUCK YOU WANT TO. + diff --git a/cmd/tomlv/README.md b/cmd/tomlv/README.md new file mode 100644 index 000000000..5df0dc32b --- /dev/null +++ b/cmd/tomlv/README.md @@ -0,0 +1,22 @@ +# TOML Validator + +If Go is installed, it's simple to try it out: + +```bash +go get github.com/BurntSushi/toml/cmd/tomlv +tomlv some-toml-file.toml +``` + +You can see the types of every key in a TOML file with: + +```bash +tomlv -types some-toml-file.toml +``` + +At the moment, only one error message is reported at a time. Error messages +include line numbers. No output means that the files given are valid TOML, or +there is a bug in `tomlv`. + +Compatible with TOML version +[v0.1.0](https://github.com/mojombo/toml/blob/master/versions/toml-v0.1.0.md) + diff --git a/cmd/tomlv/main.go b/cmd/tomlv/main.go new file mode 100644 index 000000000..c7d689a7e --- /dev/null +++ b/cmd/tomlv/main.go @@ -0,0 +1,61 @@ +// Command tomlv validates TOML documents and prints each key's type. +package main + +import ( + "flag" + "fmt" + "log" + "os" + "path" + "strings" + "text/tabwriter" + + "github.com/BurntSushi/toml" +) + +var ( + flagTypes = false +) + +func init() { + log.SetFlags(0) + + flag.BoolVar(&flagTypes, "types", flagTypes, + "When set, the types of every defined key will be shown.") + + flag.Usage = usage + flag.Parse() +} + +func usage() { + log.Printf("Usage: %s toml-file [ toml-file ... ]\n", + path.Base(os.Args[0])) + flag.PrintDefaults() + + os.Exit(1) +} + +func main() { + if flag.NArg() < 1 { + flag.Usage() + } + for _, f := range flag.Args() { + var tmp interface{} + md, err := toml.DecodeFile(f, &tmp) + if err != nil { + log.Fatalf("Error in '%s': %s", f, err) + } + if flagTypes { + printTypes(md) + } + } +} + +func printTypes(md toml.MetaData) { + tabw := tabwriter.NewWriter(os.Stdout, 0, 0, 2, ' ', 0) + for _, key := range md.Keys() { + fmt.Fprintf(tabw, "%s%s\t%s\n", + strings.Repeat(" ", len(key)-1), key, md.Type(key...)) + } + tabw.Flush() +} diff --git a/decode.go b/decode.go new file mode 100644 index 000000000..b6d75d042 --- /dev/null +++ b/decode.go @@ -0,0 +1,472 @@ +package toml + +import ( + "fmt" + "io" + "io/ioutil" + "math" + "reflect" + "strings" + "time" +) + +var e = fmt.Errorf + +// Primitive is a TOML value that hasn't been decoded into a Go value. +// When using the various `Decode*` functions, the type `Primitive` may +// be given to any value, and its decoding will be delayed. +// +// A `Primitive` value can be decoded using the `PrimitiveDecode` function. +// +// The underlying representation of a `Primitive` value is subject to change. +// Do not rely on it. +// +// N.B. Primitive values are still parsed, so using them will only avoid +// the overhead of reflection. They can be useful when you don't know the +// exact type of TOML data until run time. +type Primitive struct { + undecoded interface{} + context Key +} + +// DEPRECATED! +// +// Use MetaData.PrimitiveDecode instead. +func PrimitiveDecode(primValue Primitive, v interface{}) error { + md := MetaData{decoded: make(map[string]bool)} + return md.unify(primValue.undecoded, rvalue(v)) +} + +// PrimitiveDecode is just like the other `Decode*` functions, except it +// decodes a TOML value that has already been parsed. Valid primitive values +// can *only* be obtained from values filled by the decoder functions, +// including this method. (i.e., `v` may contain more `Primitive` +// values.) +// +// Meta data for primitive values is included in the meta data returned by +// the `Decode*` functions with one exception: keys returned by the Undecoded +// method will only reflect keys that were decoded. Namely, any keys hidden +// behind a Primitive will be considered undecoded. Executing this method will +// update the undecoded keys in the meta data. (See the example.) +func (md *MetaData) PrimitiveDecode(primValue Primitive, v interface{}) error { + md.context = primValue.context + defer func() { md.context = nil }() + return md.unify(primValue.undecoded, rvalue(v)) +} + +// Decode will decode the contents of `data` in TOML format into a pointer +// `v`. +// +// TOML hashes correspond to Go structs or maps. (Dealer's choice. They can be +// used interchangeably.) +// +// TOML arrays of tables correspond to either a slice of structs or a slice +// of maps. +// +// TOML datetimes correspond to Go `time.Time` values. +// +// All other TOML types (float, string, int, bool and array) correspond +// to the obvious Go types. +// +// An exception to the above rules is if a type implements the +// encoding.TextUnmarshaler interface. In this case, any primitive TOML value +// (floats, strings, integers, booleans and datetimes) will be converted to +// a byte string and given to the value's UnmarshalText method. See the +// Unmarshaler example for a demonstration with time duration strings. +// +// Key mapping +// +// TOML keys can map to either keys in a Go map or field names in a Go +// struct. The special `toml` struct tag may be used to map TOML keys to +// struct fields that don't match the key name exactly. (See the example.) +// A case insensitive match to struct names will be tried if an exact match +// can't be found. +// +// The mapping between TOML values and Go values is loose. That is, there +// may exist TOML values that cannot be placed into your representation, and +// there may be parts of your representation that do not correspond to +// TOML values. This loose mapping can be made stricter by using the IsDefined +// and/or Undecoded methods on the MetaData returned. +// +// This decoder will not handle cyclic types. If a cyclic type is passed, +// `Decode` will not terminate. +func Decode(data string, v interface{}) (MetaData, error) { + p, err := parse(data) + if err != nil { + return MetaData{}, err + } + md := MetaData{ + p.mapping, p.types, p.ordered, + make(map[string]bool, len(p.ordered)), nil, + } + return md, md.unify(p.mapping, rvalue(v)) +} + +// DecodeFile is just like Decode, except it will automatically read the +// contents of the file at `fpath` and decode it for you. +func DecodeFile(fpath string, v interface{}) (MetaData, error) { + bs, err := ioutil.ReadFile(fpath) + if err != nil { + return MetaData{}, err + } + return Decode(string(bs), v) +} + +// DecodeReader is just like Decode, except it will consume all bytes +// from the reader and decode it for you. +func DecodeReader(r io.Reader, v interface{}) (MetaData, error) { + bs, err := ioutil.ReadAll(r) + if err != nil { + return MetaData{}, err + } + return Decode(string(bs), v) +} + +// unify performs a sort of type unification based on the structure of `rv`, +// which is the client representation. +// +// Any type mismatch produces an error. Finding a type that we don't know +// how to handle produces an unsupported type error. +func (md *MetaData) unify(data interface{}, rv reflect.Value) error { + // Special case. Look for a `Primitive` value. + if rv.Type() == reflect.TypeOf((*Primitive)(nil)).Elem() { + // Save the undecoded data and the key context into the primitive + // value. + context := make(Key, len(md.context)) + copy(context, md.context) + rv.Set(reflect.ValueOf(Primitive{ + undecoded: data, + context: context, + })) + return nil + } + + // Special case. Handle time.Time values specifically. + // TODO: Remove this code when we decide to drop support for Go 1.1. + // This isn't necessary in Go 1.2 because time.Time satisfies the encoding + // interfaces. + if rv.Type().AssignableTo(rvalue(time.Time{}).Type()) { + return md.unifyDatetime(data, rv) + } + + // Special case. Look for a value satisfying the TextUnmarshaler interface. + if v, ok := rv.Interface().(TextUnmarshaler); ok { + return md.unifyText(data, v) + } + // BUG(burntsushi) + // The behavior here is incorrect whenever a Go type satisfies the + // encoding.TextUnmarshaler interface but also corresponds to a TOML + // hash or array. In particular, the unmarshaler should only be applied + // to primitive TOML values. But at this point, it will be applied to + // all kinds of values and produce an incorrect error whenever those values + // are hashes or arrays (including arrays of tables). + + k := rv.Kind() + + // laziness + if k >= reflect.Int && k <= reflect.Uint64 { + return md.unifyInt(data, rv) + } + switch k { + case reflect.Ptr: + elem := reflect.New(rv.Type().Elem()) + err := md.unify(data, reflect.Indirect(elem)) + if err != nil { + return err + } + rv.Set(elem) + return nil + case reflect.Struct: + return md.unifyStruct(data, rv) + case reflect.Map: + return md.unifyMap(data, rv) + case reflect.Array: + return md.unifyArray(data, rv) + case reflect.Slice: + return md.unifySlice(data, rv) + case reflect.String: + return md.unifyString(data, rv) + case reflect.Bool: + return md.unifyBool(data, rv) + case reflect.Interface: + // we only support empty interfaces. + if rv.NumMethod() > 0 { + return e("Unsupported type '%s'.", rv.Kind()) + } + return md.unifyAnything(data, rv) + case reflect.Float32: + fallthrough + case reflect.Float64: + return md.unifyFloat64(data, rv) + } + return e("Unsupported type '%s'.", rv.Kind()) +} + +func (md *MetaData) unifyStruct(mapping interface{}, rv reflect.Value) error { + tmap, ok := mapping.(map[string]interface{}) + if !ok { + return mismatch(rv, "map", mapping) + } + + for key, datum := range tmap { + var f *field + fields := cachedTypeFields(rv.Type()) + for i := range fields { + ff := &fields[i] + if ff.name == key { + f = ff + break + } + if f == nil && strings.EqualFold(ff.name, key) { + f = ff + } + } + if f != nil { + subv := rv + for _, i := range f.index { + subv = indirect(subv.Field(i)) + } + if isUnifiable(subv) { + md.decoded[md.context.add(key).String()] = true + md.context = append(md.context, key) + if err := md.unify(datum, subv); err != nil { + return e("Type mismatch for '%s.%s': %s", + rv.Type().String(), f.name, err) + } + md.context = md.context[0 : len(md.context)-1] + } else if f.name != "" { + // Bad user! No soup for you! + return e("Field '%s.%s' is unexported, and therefore cannot "+ + "be loaded with reflection.", rv.Type().String(), f.name) + } + } + } + return nil +} + +func (md *MetaData) unifyMap(mapping interface{}, rv reflect.Value) error { + tmap, ok := mapping.(map[string]interface{}) + if !ok { + return badtype("map", mapping) + } + if rv.IsNil() { + rv.Set(reflect.MakeMap(rv.Type())) + } + for k, v := range tmap { + md.decoded[md.context.add(k).String()] = true + md.context = append(md.context, k) + + rvkey := indirect(reflect.New(rv.Type().Key())) + rvval := reflect.Indirect(reflect.New(rv.Type().Elem())) + if err := md.unify(v, rvval); err != nil { + return err + } + md.context = md.context[0 : len(md.context)-1] + + rvkey.SetString(k) + rv.SetMapIndex(rvkey, rvval) + } + return nil +} + +func (md *MetaData) unifyArray(data interface{}, rv reflect.Value) error { + datav := reflect.ValueOf(data) + if datav.Kind() != reflect.Slice { + return badtype("slice", data) + } + sliceLen := datav.Len() + if sliceLen != rv.Len() { + return e("expected array length %d; got TOML array of length %d", + rv.Len(), sliceLen) + } + return md.unifySliceArray(datav, rv) +} + +func (md *MetaData) unifySlice(data interface{}, rv reflect.Value) error { + datav := reflect.ValueOf(data) + if datav.Kind() != reflect.Slice { + return badtype("slice", data) + } + sliceLen := datav.Len() + if rv.IsNil() { + rv.Set(reflect.MakeSlice(rv.Type(), sliceLen, sliceLen)) + } + return md.unifySliceArray(datav, rv) +} + +func (md *MetaData) unifySliceArray(data, rv reflect.Value) error { + sliceLen := data.Len() + for i := 0; i < sliceLen; i++ { + v := data.Index(i).Interface() + sliceval := indirect(rv.Index(i)) + if err := md.unify(v, sliceval); err != nil { + return err + } + } + return nil +} + +func (md *MetaData) unifyDatetime(data interface{}, rv reflect.Value) error { + if _, ok := data.(time.Time); ok { + rv.Set(reflect.ValueOf(data)) + return nil + } + return badtype("time.Time", data) +} + +func (md *MetaData) unifyString(data interface{}, rv reflect.Value) error { + if s, ok := data.(string); ok { + rv.SetString(s) + return nil + } + return badtype("string", data) +} + +func (md *MetaData) unifyFloat64(data interface{}, rv reflect.Value) error { + if num, ok := data.(float64); ok { + switch rv.Kind() { + case reflect.Float32: + fallthrough + case reflect.Float64: + rv.SetFloat(num) + default: + panic("bug") + } + return nil + } + return badtype("float", data) +} + +func (md *MetaData) unifyInt(data interface{}, rv reflect.Value) error { + if num, ok := data.(int64); ok { + if rv.Kind() >= reflect.Int && rv.Kind() <= reflect.Int64 { + switch rv.Kind() { + case reflect.Int, reflect.Int64: + // No bounds checking necessary. + case reflect.Int8: + if num < math.MinInt8 || num > math.MaxInt8 { + return e("Value '%d' is out of range for int8.", num) + } + case reflect.Int16: + if num < math.MinInt16 || num > math.MaxInt16 { + return e("Value '%d' is out of range for int16.", num) + } + case reflect.Int32: + if num < math.MinInt32 || num > math.MaxInt32 { + return e("Value '%d' is out of range for int32.", num) + } + } + rv.SetInt(num) + } else if rv.Kind() >= reflect.Uint && rv.Kind() <= reflect.Uint64 { + unum := uint64(num) + switch rv.Kind() { + case reflect.Uint, reflect.Uint64: + // No bounds checking necessary. + case reflect.Uint8: + if num < 0 || unum > math.MaxUint8 { + return e("Value '%d' is out of range for uint8.", num) + } + case reflect.Uint16: + if num < 0 || unum > math.MaxUint16 { + return e("Value '%d' is out of range for uint16.", num) + } + case reflect.Uint32: + if num < 0 || unum > math.MaxUint32 { + return e("Value '%d' is out of range for uint32.", num) + } + } + rv.SetUint(unum) + } else { + panic("unreachable") + } + return nil + } + return badtype("integer", data) +} + +func (md *MetaData) unifyBool(data interface{}, rv reflect.Value) error { + if b, ok := data.(bool); ok { + rv.SetBool(b) + return nil + } + return badtype("boolean", data) +} + +func (md *MetaData) unifyAnything(data interface{}, rv reflect.Value) error { + rv.Set(reflect.ValueOf(data)) + return nil +} + +func (md *MetaData) unifyText(data interface{}, v TextUnmarshaler) error { + var s string + switch sdata := data.(type) { + case TextMarshaler: + text, err := sdata.MarshalText() + if err != nil { + return err + } + s = string(text) + case fmt.Stringer: + s = sdata.String() + case string: + s = sdata + case bool: + s = fmt.Sprintf("%v", sdata) + case int64: + s = fmt.Sprintf("%d", sdata) + case float64: + s = fmt.Sprintf("%f", sdata) + default: + return badtype("primitive (string-like)", data) + } + if err := v.UnmarshalText([]byte(s)); err != nil { + return err + } + return nil +} + +// rvalue returns a reflect.Value of `v`. All pointers are resolved. +func rvalue(v interface{}) reflect.Value { + return indirect(reflect.ValueOf(v)) +} + +// indirect returns the value pointed to by a pointer. +// Pointers are followed until the value is not a pointer. +// New values are allocated for each nil pointer. +// +// An exception to this rule is if the value satisfies an interface of +// interest to us (like encoding.TextUnmarshaler). +func indirect(v reflect.Value) reflect.Value { + if v.Kind() != reflect.Ptr { + if v.CanAddr() { + pv := v.Addr() + if _, ok := pv.Interface().(TextUnmarshaler); ok { + return pv + } + } + return v + } + if v.IsNil() { + v.Set(reflect.New(v.Type().Elem())) + } + return indirect(reflect.Indirect(v)) +} + +func isUnifiable(rv reflect.Value) bool { + if rv.CanSet() { + return true + } + if _, ok := rv.Interface().(TextUnmarshaler); ok { + return true + } + return false +} + +func badtype(expected string, data interface{}) error { + return e("Expected %s but found '%T'.", expected, data) +} + +func mismatch(user reflect.Value, expected string, data interface{}) error { + return e("Type mismatch for %s. Expected %s but found '%T'.", + user.Type().String(), expected, data) +} diff --git a/decode_meta.go b/decode_meta.go new file mode 100644 index 000000000..c8114453b --- /dev/null +++ b/decode_meta.go @@ -0,0 +1,99 @@ +package toml + +import "strings" + +// MetaData allows access to meta information about TOML data that may not +// be inferrable via reflection. In particular, whether a key has been defined +// and the TOML type of a key. +type MetaData struct { + mapping map[string]interface{} + types map[string]tomlType + keys []Key + decoded map[string]bool + context Key // Used only during decoding. +} + +// IsDefined returns true if the key given exists in the TOML data. The key +// should be specified hierarchially. e.g., +// +// // access the TOML key 'a.b.c' +// IsDefined("a", "b", "c") +// +// IsDefined will return false if an empty key given. Keys are case sensitive. +func (md *MetaData) IsDefined(key ...string) bool { + if len(key) == 0 { + return false + } + + var hash map[string]interface{} + var ok bool + var hashOrVal interface{} = md.mapping + for _, k := range key { + if hash, ok = hashOrVal.(map[string]interface{}); !ok { + return false + } + if hashOrVal, ok = hash[k]; !ok { + return false + } + } + return true +} + +// Type returns a string representation of the type of the key specified. +// +// Type will return the empty string if given an empty key or a key that +// does not exist. Keys are case sensitive. +func (md *MetaData) Type(key ...string) string { + fullkey := strings.Join(key, ".") + if typ, ok := md.types[fullkey]; ok { + return typ.typeString() + } + return "" +} + +// Key is the type of any TOML key, including key groups. Use (MetaData).Keys +// to get values of this type. +type Key []string + +func (k Key) String() string { + return strings.Join(k, ".") +} + +func (k Key) add(piece string) Key { + newKey := make(Key, len(k)+1) + copy(newKey, k) + newKey[len(k)] = piece + return newKey +} + +// Keys returns a slice of every key in the TOML data, including key groups. +// Each key is itself a slice, where the first element is the top of the +// hierarchy and the last is the most specific. +// +// The list will have the same order as the keys appeared in the TOML data. +// +// All keys returned are non-empty. +func (md *MetaData) Keys() []Key { + return md.keys +} + +// Undecoded returns all keys that have not been decoded in the order in which +// they appear in the original TOML document. +// +// This includes keys that haven't been decoded because of a Primitive value. +// Once the Primitive value is decoded, the keys will be considered decoded. +// +// Also note that decoding into an empty interface will result in no decoding, +// and so no keys will be considered decoded. +// +// In this sense, the Undecoded keys correspond to keys in the TOML document +// that do not have a concrete type in your representation. +func (md *MetaData) Undecoded() []Key { + undecoded := make([]Key, 0, len(md.keys)) + for _, key := range md.keys { + if !md.decoded[key.String()] { + undecoded = append(undecoded, key) + } + } + return undecoded +} diff --git a/decode_test.go b/decode_test.go new file mode 100644 index 000000000..b940333dc --- /dev/null +++ b/decode_test.go @@ -0,0 +1,540 @@ +package toml + +import ( + "fmt" + "log" + "reflect" + "testing" + "time" +) + +func init() { + log.SetFlags(0) +} + +func TestDecodeSimple(t *testing.T) { + var testSimple = ` +age = 250 +andrew = "gallant" +kait = "brady" +now = 1987-07-05T05:45:00Z +yesOrNo = true +pi = 3.14 +colors = [ + ["red", "green", "blue"], + ["cyan", "magenta", "yellow", "black"], +] + +[My.Cats] +plato = "cat 1" +cauchy = "cat 2" +` + + type cats struct { + Plato string + Cauchy string + } + type simple struct { + Age int + Colors [][]string + Pi float64 + YesOrNo bool + Now time.Time + Andrew string + Kait string + My map[string]cats + } + + var val simple + _, err := Decode(testSimple, &val) + if err != nil { + t.Fatal(err) + } + + now, err := time.Parse("2006-01-02T15:04:05", "1987-07-05T05:45:00") + if err != nil { + panic(err) + } + var answer = simple{ + Age: 250, + Andrew: "gallant", + Kait: "brady", + Now: now, + YesOrNo: true, + Pi: 3.14, + Colors: [][]string{ + {"red", "green", "blue"}, + {"cyan", "magenta", "yellow", "black"}, + }, + My: map[string]cats{ + "Cats": cats{Plato: "cat 1", Cauchy: "cat 2"}, + }, + } + if !reflect.DeepEqual(val, answer) { + t.Fatalf("Expected\n-----\n%#v\n-----\nbut got\n-----\n%#v\n", + answer, val) + } +} + +func TestDecodeEmbedded(t *testing.T) { + type Dog struct{ Name string } + type Age int + + tests := map[string]struct { + input string + decodeInto interface{} + wantDecoded interface{} + }{ + "embedded struct": { + input: `Name = "milton"`, + decodeInto: &struct{ Dog }{}, + wantDecoded: &struct{ Dog }{Dog{"milton"}}, + }, + "embedded non-nil pointer to struct": { + input: `Name = "milton"`, + decodeInto: &struct{ *Dog }{}, + wantDecoded: &struct{ *Dog }{&Dog{"milton"}}, + }, + "embedded nil pointer to struct": { + input: ``, + decodeInto: &struct{ *Dog }{}, + wantDecoded: &struct{ *Dog }{nil}, + }, + "embedded int": { + input: `Age = -5`, + decodeInto: &struct{ Age }{}, + wantDecoded: &struct{ Age }{-5}, + }, + } + + for label, test := range tests { + _, err := Decode(test.input, test.decodeInto) + if err != nil { + t.Fatal(err) + } + if !reflect.DeepEqual(test.wantDecoded, test.decodeInto) { + t.Errorf("%s: want decoded == %+v, got %+v", + label, test.wantDecoded, test.decodeInto) + } + } +} + +func TestTableArrays(t *testing.T) { + var tomlTableArrays = ` +[[albums]] +name = "Born to Run" + + [[albums.songs]] + name = "Jungleland" + + [[albums.songs]] + name = "Meeting Across the River" + +[[albums]] +name = "Born in the USA" + + [[albums.songs]] + name = "Glory Days" + + [[albums.songs]] + name = "Dancing in the Dark" +` + + type Song struct { + Name string + } + + type Album struct { + Name string + Songs []Song + } + + type Music struct { + Albums []Album + } + + expected := Music{[]Album{ + {"Born to Run", []Song{{"Jungleland"}, {"Meeting Across the River"}}}, + {"Born in the USA", []Song{{"Glory Days"}, {"Dancing in the Dark"}}}, + }} + var got Music + if _, err := Decode(tomlTableArrays, &got); err != nil { + t.Fatal(err) + } + if !reflect.DeepEqual(expected, got) { + t.Fatalf("\n%#v\n!=\n%#v\n", expected, got) + } +} + +// Case insensitive matching tests. +// A bit more comprehensive than needed given the current implementation, +// but implementations change. +// Probably still missing demonstrations of some ugly corner cases regarding +// case insensitive matching and multiple fields. +func TestCase(t *testing.T) { + var caseToml = ` +tOpString = "string" +tOpInt = 1 +tOpFloat = 1.1 +tOpBool = true +tOpdate = 2006-01-02T15:04:05Z +tOparray = [ "array" ] +Match = "i should be in Match only" +MatcH = "i should be in MatcH only" +once = "just once" +[nEst.eD] +nEstedString = "another string" +` + + type InsensitiveEd struct { + NestedString string + } + + type InsensitiveNest struct { + Ed InsensitiveEd + } + + type Insensitive struct { + TopString string + TopInt int + TopFloat float64 + TopBool bool + TopDate time.Time + TopArray []string + Match string + MatcH string + Once string + OncE string + Nest InsensitiveNest + } + + tme, err := time.Parse(time.RFC3339, time.RFC3339[:len(time.RFC3339)-5]) + if err != nil { + panic(err) + } + expected := Insensitive{ + TopString: "string", + TopInt: 1, + TopFloat: 1.1, + TopBool: true, + TopDate: tme, + TopArray: []string{"array"}, + MatcH: "i should be in MatcH only", + Match: "i should be in Match only", + Once: "just once", + OncE: "", + Nest: InsensitiveNest{ + Ed: InsensitiveEd{NestedString: "another string"}, + }, + } + var got Insensitive + if _, err := Decode(caseToml, &got); err != nil { + t.Fatal(err) + } + if !reflect.DeepEqual(expected, got) { + t.Fatalf("\n%#v\n!=\n%#v\n", expected, got) + } +} + +func TestPointers(t *testing.T) { + type Object struct { + Type string + Description string + } + + type Dict struct { + NamedObject map[string]*Object + BaseObject *Object + Strptr *string + Strptrs []*string + } + s1, s2, s3 := "blah", "abc", "def" + expected := &Dict{ + Strptr: &s1, + Strptrs: []*string{&s2, &s3}, + NamedObject: map[string]*Object{ + "foo": {"FOO", "fooooo!!!"}, + "bar": {"BAR", "ba-ba-ba-ba-barrrr!!!"}, + }, + BaseObject: &Object{"BASE", "da base"}, + } + + ex1 := ` +Strptr = "blah" +Strptrs = ["abc", "def"] + +[NamedObject.foo] +Type = "FOO" +Description = "fooooo!!!" + +[NamedObject.bar] +Type = "BAR" +Description = "ba-ba-ba-ba-barrrr!!!" + +[BaseObject] +Type = "BASE" +Description = "da base" +` + dict := new(Dict) + _, err := Decode(ex1, dict) + if err != nil { + t.Errorf("Decode error: %v", err) + } + if !reflect.DeepEqual(expected, dict) { + t.Fatalf("\n%#v\n!=\n%#v\n", expected, dict) + } +} + +type sphere struct { + Center [3]float64 + Radius float64 +} + +func TestDecodeSimpleArray(t *testing.T) { + var s1 sphere + if _, err := Decode(`center = [0.0, 1.5, 0.0]`, &s1); err != nil { + t.Fatal(err) + } +} + +func TestDecodeArrayWrongSize(t *testing.T) { + var s1 sphere + if _, err := Decode(`center = [0.1, 2.3]`, &s1); err == nil { + t.Fatal("Expected array type mismatch error") + } +} + +func TestDecodeLargeIntoSmallInt(t *testing.T) { + type table struct { + Value int8 + } + var tab table + if _, err := Decode(`value = 500`, &tab); err == nil { + t.Fatal("Expected integer out-of-bounds error.") + } +} + +func TestDecodeSizedInts(t *testing.T) { + type table struct { + U8 uint8 + U16 uint16 + U32 uint32 + U64 uint64 + U uint + I8 int8 + I16 int16 + I32 int32 + I64 int64 + I int + } + answer := table{1, 1, 1, 1, 1, -1, -1, -1, -1, -1} + toml := ` + u8 = 1 + u16 = 1 + u32 = 1 + u64 = 1 + u = 1 + i8 = -1 + i16 = -1 + i32 = -1 + i64 = -1 + i = -1 + ` + var tab table + if _, err := Decode(toml, &tab); err != nil { + t.Fatal(err.Error()) + } + if answer != tab { + t.Fatalf("Expected %#v but got %#v", answer, tab) + } +} + +func ExampleMetaData_PrimitiveDecode() { + var md MetaData + var err error + + var tomlBlob = ` +ranking = ["Springsteen", "J Geils"] + +[bands.Springsteen] +started = 1973 +albums = ["Greetings", "WIESS", "Born to Run", "Darkness"] + +[bands.J Geils] +started = 1970 +albums = ["The J. Geils Band", "Full House", "Blow Your Face Out"] +` + + type band struct { + Started int + Albums []string + } + type classics struct { + Ranking []string + Bands map[string]Primitive + } + + // Do the initial decode. Reflection is delayed on Primitive values. + var music classics + if md, err = Decode(tomlBlob, &music); err != nil { + log.Fatal(err) + } + + // MetaData still includes information on Primitive values. + fmt.Printf("Is `bands.Springsteen` defined? %v\n", + md.IsDefined("bands", "Springsteen")) + + // Decode primitive data into Go values. + for _, artist := range music.Ranking { + // A band is a primitive value, so we need to decode it to get a + // real `band` value. + primValue := music.Bands[artist] + + var aBand band + if err = md.PrimitiveDecode(primValue, &aBand); err != nil { + log.Fatal(err) + } + fmt.Printf("%s started in %d.\n", artist, aBand.Started) + } + // Check to see if there were any fields left undecoded. + // Note that this won't be empty before decoding the Primitive value! + fmt.Printf("Undecoded: %q\n", md.Undecoded()) + + // Output: + // Is `bands.Springsteen` defined? true + // Springsteen started in 1973. + // J Geils started in 1970. + // Undecoded: [] +} + +func ExampleDecode() { + var tomlBlob = ` +# Some comments. +[alpha] +ip = "10.0.0.1" + + [alpha.config] + Ports = [ 8001, 8002 ] + Location = "Toronto" + Created = 1987-07-05T05:45:00Z + +[beta] +ip = "10.0.0.2" + + [beta.config] + Ports = [ 9001, 9002 ] + Location = "New Jersey" + Created = 1887-01-05T05:55:00Z +` + + type serverConfig struct { + Ports []int + Location string + Created time.Time + } + + type server struct { + IP string `toml:"ip"` + Config serverConfig `toml:"config"` + } + + type servers map[string]server + + var config servers + if _, err := Decode(tomlBlob, &config); err != nil { + log.Fatal(err) + } + + for _, name := range []string{"alpha", "beta"} { + s := config[name] + fmt.Printf("Server: %s (ip: %s) in %s created on %s\n", + name, s.IP, s.Config.Location, + s.Config.Created.Format("2006-01-02")) + fmt.Printf("Ports: %v\n", s.Config.Ports) + } + + // Output: + // Server: alpha (ip: 10.0.0.1) in Toronto created on 1987-07-05 + // Ports: [8001 8002] + // Server: beta (ip: 10.0.0.2) in New Jersey created on 1887-01-05 + // Ports: [9001 9002] +} + +type duration struct { + time.Duration +} + +func (d *duration) UnmarshalText(text []byte) error { + var err error + d.Duration, err = time.ParseDuration(string(text)) + return err +} + +// Example Unmarshaler shows how to decode TOML strings into your own +// custom data type. +func Example_unmarshaler() { + blob := ` +[[song]] +name = "Thunder Road" +duration = "4m49s" + +[[song]] +name = "Stairway to Heaven" +duration = "8m03s" +` + type song struct { + Name string + Duration duration + } + type songs struct { + Song []song + } + var favorites songs + if _, err := Decode(blob, &favorites); err != nil { + log.Fatal(err) + } + + // Code to implement the TextUnmarshaler interface for `duration`: + // + // type duration struct { + // time.Duration + // } + // + // func (d *duration) UnmarshalText(text []byte) error { + // var err error + // d.Duration, err = time.ParseDuration(string(text)) + // return err + // } + + for _, s := range favorites.Song { + fmt.Printf("%s (%s)\n", s.Name, s.Duration) + } + // Output: + // Thunder Road (4m49s) + // Stairway to Heaven (8m3s) +} + +// Example StrictDecoding shows how to detect whether there are keys in the +// TOML document that weren't decoded into the value given. This is useful +// for returning an error to the user if they've included extraneous fields +// in their configuration. +func Example_strictDecoding() { + var blob = ` +key1 = "value1" +key2 = "value2" +key3 = "value3" +` + type config struct { + Key1 string + Key3 string + } + + var conf config + md, err := Decode(blob, &conf) + if err != nil { + log.Fatal(err) + } + fmt.Printf("Undecoded keys: %q\n", md.Undecoded()) + // Output: + // Undecoded keys: ["key2"] +} diff --git a/doc.go b/doc.go new file mode 100644 index 000000000..fe2680004 --- /dev/null +++ b/doc.go @@ -0,0 +1,27 @@ +/* +Package toml provides facilities for decoding and encoding TOML configuration +files via reflection. There is also support for delaying decoding with +the Primitive type, and querying the set of keys in a TOML document with the +MetaData type. + +The specification implemented: https://github.com/mojombo/toml + +The sub-command github.com/BurntSushi/toml/cmd/tomlv can be used to verify +whether a file is a valid TOML document. It can also be used to print the +type of each key in a TOML document. + +Testing + +There are two important types of tests used for this package. The first is +contained inside '*_test.go' files and uses the standard Go unit testing +framework. These tests are primarily devoted to holistically testing the +decoder and encoder. + +The second type of testing is used to verify the implementation's adherence +to the TOML specification. These tests have been factored into their own +project: https://github.com/BurntSushi/toml-test + +The reason the tests are in a separate project is so that they can be used by +any implementation of TOML. Namely, it is language agnostic. +*/ +package toml diff --git a/encode.go b/encode.go new file mode 100644 index 000000000..361871347 --- /dev/null +++ b/encode.go @@ -0,0 +1,515 @@ +package toml + +import ( + "bufio" + "errors" + "fmt" + "io" + "reflect" + "sort" + "strconv" + "strings" + "time" +) + +type tomlEncodeError struct{ error } + +var ( + errArrayMixedElementTypes = errors.New( + "can't encode array with mixed element types") + errArrayNilElement = errors.New( + "can't encode array with nil element") + errNonString = errors.New( + "can't encode a map with non-string key type") + errAnonNonStruct = errors.New( + "can't encode an anonymous field that is not a struct") + errArrayNoTable = errors.New( + "TOML array element can't contain a table") + errNoKey = errors.New( + "top-level values must be a Go map or struct") + errAnything = errors.New("") // used in testing +) + +var quotedReplacer = strings.NewReplacer( + "\t", "\\t", + "\n", "\\n", + "\r", "\\r", + "\"", "\\\"", + "\\", "\\\\", +) + +// Encoder controls the encoding of Go values to a TOML document to some +// io.Writer. +// +// The indentation level can be controlled with the Indent field. +type Encoder struct { + // A single indentation level. By default it is two spaces. + Indent string + + // hasWritten is whether we have written any output to w yet. + hasWritten bool + w *bufio.Writer +} + +// NewEncoder returns a TOML encoder that encodes Go values to the io.Writer +// given. By default, a single indentation level is 2 spaces. +func NewEncoder(w io.Writer) *Encoder { + return &Encoder{ + w: bufio.NewWriter(w), + Indent: " ", + } +} + +// Encode writes a TOML representation of the Go value to the underlying +// io.Writer. If the value given cannot be encoded to a valid TOML document, +// then an error is returned. +// +// The mapping between Go values and TOML values should be precisely the same +// as for the Decode* functions. Similarly, the TextMarshaler interface is +// supported by encoding the resulting bytes as strings. (If you want to write +// arbitrary binary data then you will need to use something like base64 since +// TOML does not have any binary types.) +// +// When encoding TOML hashes (i.e., Go maps or structs), keys without any +// sub-hashes are encoded first. +// +// If a Go map is encoded, then its keys are sorted alphabetically for +// deterministic output. More control over this behavior may be provided if +// there is demand for it. +// +// Encoding Go values without a corresponding TOML representation---like map +// types with non-string keys---will cause an error to be returned. Similarly +// for mixed arrays/slices, arrays/slices with nil elements, embedded +// non-struct types and nested slices containing maps or structs. +// (e.g., [][]map[string]string is not allowed but []map[string]string is OK +// and so is []map[string][]string.) +func (enc *Encoder) Encode(v interface{}) error { + rv := eindirect(reflect.ValueOf(v)) + if err := enc.safeEncode(Key([]string{}), rv); err != nil { + return err + } + return enc.w.Flush() +} + +func (enc *Encoder) safeEncode(key Key, rv reflect.Value) (err error) { + defer func() { + if r := recover(); r != nil { + if terr, ok := r.(tomlEncodeError); ok { + err = terr.error + return + } + panic(r) + } + }() + enc.encode(key, rv) + return nil +} + +func (enc *Encoder) encode(key Key, rv reflect.Value) { + // Special case. Time needs to be in ISO8601 format. + // Special case. If we can marshal the type to text, then we used that. + // Basically, this prevents the encoder for handling these types as + // generic structs (or whatever the underlying type of a TextMarshaler is). + switch rv.Interface().(type) { + case time.Time, TextMarshaler: + enc.keyEqElement(key, rv) + return + } + + k := rv.Kind() + switch k { + case reflect.Int, reflect.Int8, reflect.Int16, reflect.Int32, reflect.Int64, + reflect.Uint, reflect.Uint8, reflect.Uint16, reflect.Uint32, + reflect.Uint64, + reflect.Float32, reflect.Float64, reflect.String, reflect.Bool: + enc.keyEqElement(key, rv) + case reflect.Array, reflect.Slice: + if typeEqual(tomlArrayHash, tomlTypeOfGo(rv)) { + enc.eArrayOfTables(key, rv) + } else { + enc.keyEqElement(key, rv) + } + case reflect.Interface: + if rv.IsNil() { + return + } + enc.encode(key, rv.Elem()) + case reflect.Map: + if rv.IsNil() { + return + } + enc.eTable(key, rv) + case reflect.Ptr: + if rv.IsNil() { + return + } + enc.encode(key, rv.Elem()) + case reflect.Struct: + enc.eTable(key, rv) + default: + panic(e("Unsupported type for key '%s': %s", key, k)) + } +} + +// eElement encodes any value that can be an array element (primitives and +// arrays). +func (enc *Encoder) eElement(rv reflect.Value) { + switch v := rv.Interface().(type) { + case time.Time: + // Special case time.Time as a primitive. Has to come before + // TextMarshaler below because time.Time implements + // encoding.TextMarshaler, but we need to always use UTC. + enc.wf(v.In(time.FixedZone("UTC", 0)).Format("2006-01-02T15:04:05Z")) + return + case TextMarshaler: + // Special case. Use text marshaler if it's available for this value. + if s, err := v.MarshalText(); err != nil { + encPanic(err) + } else { + enc.writeQuoted(string(s)) + } + return + } + switch rv.Kind() { + case reflect.Bool: + enc.wf(strconv.FormatBool(rv.Bool())) + case reflect.Int, reflect.Int8, reflect.Int16, reflect.Int32, reflect.Int64: + enc.wf(strconv.FormatInt(rv.Int(), 10)) + case reflect.Uint, reflect.Uint8, reflect.Uint16, + reflect.Uint32, reflect.Uint64: + enc.wf(strconv.FormatUint(rv.Uint(), 10)) + case reflect.Float32: + enc.wf(floatAddDecimal(strconv.FormatFloat(rv.Float(), 'f', -1, 32))) + case reflect.Float64: + enc.wf(floatAddDecimal(strconv.FormatFloat(rv.Float(), 'f', -1, 64))) + case reflect.Array, reflect.Slice: + enc.eArrayOrSliceElement(rv) + case reflect.Interface: + enc.eElement(rv.Elem()) + case reflect.String: + enc.writeQuoted(rv.String()) + default: + panic(e("Unexpected primitive type: %s", rv.Kind())) + } +} + +// By the TOML spec, all floats must have a decimal with at least one +// number on either side. +func floatAddDecimal(fstr string) string { + if !strings.Contains(fstr, ".") { + return fstr + ".0" + } + return fstr +} + +func (enc *Encoder) writeQuoted(s string) { + enc.wf("\"%s\"", quotedReplacer.Replace(s)) +} + +func (enc *Encoder) eArrayOrSliceElement(rv reflect.Value) { + length := rv.Len() + enc.wf("[") + for i := 0; i < length; i++ { + elem := rv.Index(i) + enc.eElement(elem) + if i != length-1 { + enc.wf(", ") + } + } + enc.wf("]") +} + +func (enc *Encoder) eArrayOfTables(key Key, rv reflect.Value) { + if len(key) == 0 { + encPanic(errNoKey) + } + panicIfInvalidKey(key, true) + for i := 0; i < rv.Len(); i++ { + trv := rv.Index(i) + if isNil(trv) { + continue + } + enc.newline() + enc.wf("%s[[%s]]", enc.indentStr(key), key.String()) + enc.newline() + enc.eMapOrStruct(key, trv) + } +} + +func (enc *Encoder) eTable(key Key, rv reflect.Value) { + if len(key) == 1 { + // Output an extra new line between top-level tables. + // (The newline isn't written if nothing else has been written though.) + enc.newline() + } + if len(key) > 0 { + panicIfInvalidKey(key, true) + enc.wf("%s[%s]", enc.indentStr(key), key.String()) + enc.newline() + } + enc.eMapOrStruct(key, rv) +} + +func (enc *Encoder) eMapOrStruct(key Key, rv reflect.Value) { + switch rv := eindirect(rv); rv.Kind() { + case reflect.Map: + enc.eMap(key, rv) + case reflect.Struct: + enc.eStruct(key, rv) + default: + panic("eTable: unhandled reflect.Value Kind: " + rv.Kind().String()) + } +} + +func (enc *Encoder) eMap(key Key, rv reflect.Value) { + rt := rv.Type() + if rt.Key().Kind() != reflect.String { + encPanic(errNonString) + } + + // Sort keys so that we have deterministic output. And write keys directly + // underneath this key first, before writing sub-structs or sub-maps. + var mapKeysDirect, mapKeysSub []string + for _, mapKey := range rv.MapKeys() { + k := mapKey.String() + if typeIsHash(tomlTypeOfGo(rv.MapIndex(mapKey))) { + mapKeysSub = append(mapKeysSub, k) + } else { + mapKeysDirect = append(mapKeysDirect, k) + } + } + + var writeMapKeys = func(mapKeys []string) { + sort.Strings(mapKeys) + for _, mapKey := range mapKeys { + mrv := rv.MapIndex(reflect.ValueOf(mapKey)) + if isNil(mrv) { + // Don't write anything for nil fields. + continue + } + enc.encode(key.add(mapKey), mrv) + } + } + writeMapKeys(mapKeysDirect) + writeMapKeys(mapKeysSub) +} + +func (enc *Encoder) eStruct(key Key, rv reflect.Value) { + // Write keys for fields directly under this key first, because if we write + // a field that creates a new table, then all keys under it will be in that + // table (not the one we're writing here). + rt := rv.Type() + var fieldsDirect, fieldsSub [][]int + var addFields func(rt reflect.Type, rv reflect.Value, start []int) + addFields = func(rt reflect.Type, rv reflect.Value, start []int) { + for i := 0; i < rt.NumField(); i++ { + f := rt.Field(i) + // skip unexporded fields + if f.PkgPath != "" { + continue + } + frv := rv.Field(i) + if f.Anonymous { + frv := eindirect(frv) + t := frv.Type() + if t.Kind() != reflect.Struct { + encPanic(errAnonNonStruct) + } + addFields(t, frv, f.Index) + } else if typeIsHash(tomlTypeOfGo(frv)) { + fieldsSub = append(fieldsSub, append(start, f.Index...)) + } else { + fieldsDirect = append(fieldsDirect, append(start, f.Index...)) + } + } + } + addFields(rt, rv, nil) + + var writeFields = func(fields [][]int) { + for _, fieldIndex := range fields { + sft := rt.FieldByIndex(fieldIndex) + sf := rv.FieldByIndex(fieldIndex) + if isNil(sf) { + // Don't write anything for nil fields. + continue + } + + keyName := sft.Tag.Get("toml") + if keyName == "-" { + continue + } + if keyName == "" { + keyName = sft.Name + } + enc.encode(key.add(keyName), sf) + } + } + writeFields(fieldsDirect) + writeFields(fieldsSub) +} + +// tomlTypeName returns the TOML type name of the Go value's type. It is used to +// determine whether the types of array elements are mixed (which is forbidden). +// If the Go value is nil, then it is illegal for it to be an array element, and +// valueIsNil is returned as true. + +// Returns the TOML type of a Go value. The type may be `nil`, which means +// no concrete TOML type could be found. +func tomlTypeOfGo(rv reflect.Value) tomlType { + if isNil(rv) || !rv.IsValid() { + return nil + } + switch rv.Kind() { + case reflect.Bool: + return tomlBool + case reflect.Int, reflect.Int8, reflect.Int16, reflect.Int32, reflect.Int64, + reflect.Uint, reflect.Uint8, reflect.Uint16, reflect.Uint32, + reflect.Uint64: + return tomlInteger + case reflect.Float32, reflect.Float64: + return tomlFloat + case reflect.Array, reflect.Slice: + if typeEqual(tomlHash, tomlArrayType(rv)) { + return tomlArrayHash + } else { + return tomlArray + } + case reflect.Ptr, reflect.Interface: + return tomlTypeOfGo(rv.Elem()) + case reflect.String: + return tomlString + case reflect.Map: + return tomlHash + case reflect.Struct: + switch rv.Interface().(type) { + case time.Time: + return tomlDatetime + case TextMarshaler: + return tomlString + default: + return tomlHash + } + default: + panic("unexpected reflect.Kind: " + rv.Kind().String()) + } +} + +// tomlArrayType returns the element type of a TOML array. The type returned +// may be nil if it cannot be determined (e.g., a nil slice or a zero length +// slize). This function may also panic if it finds a type that cannot be +// expressed in TOML (such as nil elements, heterogeneous arrays or directly +// nested arrays of tables). +func tomlArrayType(rv reflect.Value) tomlType { + if isNil(rv) || !rv.IsValid() || rv.Len() == 0 { + return nil + } + firstType := tomlTypeOfGo(rv.Index(0)) + if firstType == nil { + encPanic(errArrayNilElement) + } + + rvlen := rv.Len() + for i := 1; i < rvlen; i++ { + elem := rv.Index(i) + switch elemType := tomlTypeOfGo(elem); { + case elemType == nil: + encPanic(errArrayNilElement) + case !typeEqual(firstType, elemType): + encPanic(errArrayMixedElementTypes) + } + } + // If we have a nested array, then we must make sure that the nested + // array contains ONLY primitives. + // This checks arbitrarily nested arrays. + if typeEqual(firstType, tomlArray) || typeEqual(firstType, tomlArrayHash) { + nest := tomlArrayType(eindirect(rv.Index(0))) + if typeEqual(nest, tomlHash) || typeEqual(nest, tomlArrayHash) { + encPanic(errArrayNoTable) + } + } + return firstType +} + +func (enc *Encoder) newline() { + if enc.hasWritten { + enc.wf("\n") + } +} + +func (enc *Encoder) keyEqElement(key Key, val reflect.Value) { + if len(key) == 0 { + encPanic(errNoKey) + } + panicIfInvalidKey(key, false) + enc.wf("%s%s = ", enc.indentStr(key), key[len(key)-1]) + enc.eElement(val) + enc.newline() +} + +func (enc *Encoder) wf(format string, v ...interface{}) { + if _, err := fmt.Fprintf(enc.w, format, v...); err != nil { + encPanic(err) + } + enc.hasWritten = true +} + +func (enc *Encoder) indentStr(key Key) string { + return strings.Repeat(enc.Indent, len(key)-1) +} + +func encPanic(err error) { + panic(tomlEncodeError{err}) +} + +func eindirect(v reflect.Value) reflect.Value { + switch v.Kind() { + case reflect.Ptr, reflect.Interface: + return eindirect(v.Elem()) + default: + return v + } +} + +func isNil(rv reflect.Value) bool { + switch rv.Kind() { + case reflect.Interface, reflect.Map, reflect.Ptr, reflect.Slice: + return rv.IsNil() + default: + return false + } +} + +func panicIfInvalidKey(key Key, hash bool) { + if hash { + for _, k := range key { + if !isValidTableName(k) { + encPanic(e("Key '%s' is not a valid table name. Table names "+ + "cannot contain '[', ']' or '.'.", key.String())) + } + } + } else { + if !isValidKeyName(key[len(key)-1]) { + encPanic(e("Key '%s' is not a name. Key names "+ + "cannot contain whitespace.", key.String())) + } + } +} + +func isValidTableName(s string) bool { + if len(s) == 0 { + return false + } + for _, r := range s { + if r == '[' || r == ']' || r == '.' { + return false + } + } + return true +} + +func isValidKeyName(s string) bool { + if len(s) == 0 { + return false + } + return true +} diff --git a/encode_test.go b/encode_test.go new file mode 100644 index 000000000..74a5ee5d2 --- /dev/null +++ b/encode_test.go @@ -0,0 +1,506 @@ +package toml + +import ( + "bytes" + "fmt" + "log" + "net" + "testing" + "time" +) + +func TestEncodeRoundTrip(t *testing.T) { + type Config struct { + Age int + Cats []string + Pi float64 + Perfection []int + DOB time.Time + Ipaddress net.IP + } + + var inputs = Config{ + 13, + []string{"one", "two", "three"}, + 3.145, + []int{11, 2, 3, 4}, + time.Now(), + net.ParseIP("192.168.59.254"), + } + + var firstBuffer bytes.Buffer + e := NewEncoder(&firstBuffer) + err := e.Encode(inputs) + if err != nil { + t.Fatal(err) + } + var outputs Config + if _, err := Decode(firstBuffer.String(), &outputs); err != nil { + log.Printf("Could not decode:\n-----\n%s\n-----\n", + firstBuffer.String()) + t.Fatal(err) + } + + // could test each value individually, but I'm lazy + var secondBuffer bytes.Buffer + e2 := NewEncoder(&secondBuffer) + err = e2.Encode(outputs) + if err != nil { + t.Fatal(err) + } + if firstBuffer.String() != secondBuffer.String() { + t.Error( + firstBuffer.String(), + "\n\n is not identical to\n\n", + secondBuffer.String()) + } +} + +// XXX(burntsushi) +// I think these tests probably should be removed. They are good, but they +// ought to be obsolete by toml-test. +func TestEncode(t *testing.T) { + type Embedded struct { + Int int `toml:"_int"` + } + type NonStruct int + + date := time.Date(2014, 5, 11, 20, 30, 40, 0, time.FixedZone("IST", 3600)) + dateStr := "2014-05-11T19:30:40Z" + + tests := map[string]struct { + input interface{} + wantOutput string + wantError error + }{ + "bool field": { + input: struct { + BoolTrue bool + BoolFalse bool + }{true, false}, + wantOutput: "BoolTrue = true\nBoolFalse = false\n", + }, + "int fields": { + input: struct { + Int int + Int8 int8 + Int16 int16 + Int32 int32 + Int64 int64 + }{1, 2, 3, 4, 5}, + wantOutput: "Int = 1\nInt8 = 2\nInt16 = 3\nInt32 = 4\nInt64 = 5\n", + }, + "uint fields": { + input: struct { + Uint uint + Uint8 uint8 + Uint16 uint16 + Uint32 uint32 + Uint64 uint64 + }{1, 2, 3, 4, 5}, + wantOutput: "Uint = 1\nUint8 = 2\nUint16 = 3\nUint32 = 4" + + "\nUint64 = 5\n", + }, + "float fields": { + input: struct { + Float32 float32 + Float64 float64 + }{1.5, 2.5}, + wantOutput: "Float32 = 1.5\nFloat64 = 2.5\n", + }, + "string field": { + input: struct{ String string }{"foo"}, + wantOutput: "String = \"foo\"\n", + }, + "string field and unexported field": { + input: struct { + String string + unexported int + }{"foo", 0}, + wantOutput: "String = \"foo\"\n", + }, + "datetime field in UTC": { + input: struct{ Date time.Time }{date}, + wantOutput: fmt.Sprintf("Date = %s\n", dateStr), + }, + "datetime field as primitive": { + // Using a map here to fail if isStructOrMap() returns true for + // time.Time. + input: map[string]interface{}{ + "Date": date, + "Int": 1, + }, + wantOutput: fmt.Sprintf("Date = %s\nInt = 1\n", dateStr), + }, + "array fields": { + input: struct { + IntArray0 [0]int + IntArray3 [3]int + }{[0]int{}, [3]int{1, 2, 3}}, + wantOutput: "IntArray0 = []\nIntArray3 = [1, 2, 3]\n", + }, + "slice fields": { + input: struct{ IntSliceNil, IntSlice0, IntSlice3 []int }{ + nil, []int{}, []int{1, 2, 3}, + }, + wantOutput: "IntSlice0 = []\nIntSlice3 = [1, 2, 3]\n", + }, + "datetime slices": { + input: struct{ DatetimeSlice []time.Time }{ + []time.Time{date, date}, + }, + wantOutput: fmt.Sprintf("DatetimeSlice = [%s, %s]\n", + dateStr, dateStr), + }, + "nested arrays and slices": { + input: struct { + SliceOfArrays [][2]int + ArrayOfSlices [2][]int + SliceOfArraysOfSlices [][2][]int + ArrayOfSlicesOfArrays [2][][2]int + SliceOfMixedArrays [][2]interface{} + ArrayOfMixedSlices [2][]interface{} + }{ + [][2]int{{1, 2}, {3, 4}}, + [2][]int{{1, 2}, {3, 4}}, + [][2][]int{ + { + {1, 2}, {3, 4}, + }, + { + {5, 6}, {7, 8}, + }, + }, + [2][][2]int{ + { + {1, 2}, {3, 4}, + }, + { + {5, 6}, {7, 8}, + }, + }, + [][2]interface{}{ + {1, 2}, {"a", "b"}, + }, + [2][]interface{}{ + {1, 2}, {"a", "b"}, + }, + }, + wantOutput: `SliceOfArrays = [[1, 2], [3, 4]] +ArrayOfSlices = [[1, 2], [3, 4]] +SliceOfArraysOfSlices = [[[1, 2], [3, 4]], [[5, 6], [7, 8]]] +ArrayOfSlicesOfArrays = [[[1, 2], [3, 4]], [[5, 6], [7, 8]]] +SliceOfMixedArrays = [[1, 2], ["a", "b"]] +ArrayOfMixedSlices = [[1, 2], ["a", "b"]] +`, + }, + "empty slice": { + input: struct{ Empty []interface{} }{[]interface{}{}}, + wantOutput: "Empty = []\n", + }, + "(error) slice with element type mismatch (string and integer)": { + input: struct{ Mixed []interface{} }{[]interface{}{1, "a"}}, + wantError: errArrayMixedElementTypes, + }, + "(error) slice with element type mismatch (integer and float)": { + input: struct{ Mixed []interface{} }{[]interface{}{1, 2.5}}, + wantError: errArrayMixedElementTypes, + }, + "slice with elems of differing Go types, same TOML types": { + input: struct { + MixedInts []interface{} + MixedFloats []interface{} + }{ + []interface{}{ + int(1), int8(2), int16(3), int32(4), int64(5), + uint(1), uint8(2), uint16(3), uint32(4), uint64(5), + }, + []interface{}{float32(1.5), float64(2.5)}, + }, + wantOutput: "MixedInts = [1, 2, 3, 4, 5, 1, 2, 3, 4, 5]\n" + + "MixedFloats = [1.5, 2.5]\n", + }, + "(error) slice w/ element type mismatch (one is nested array)": { + input: struct{ Mixed []interface{} }{ + []interface{}{1, []interface{}{2}}, + }, + wantError: errArrayMixedElementTypes, + }, + "(error) slice with 1 nil element": { + input: struct{ NilElement1 []interface{} }{[]interface{}{nil}}, + wantError: errArrayNilElement, + }, + "(error) slice with 1 nil element (and other non-nil elements)": { + input: struct{ NilElement []interface{} }{ + []interface{}{1, nil}, + }, + wantError: errArrayNilElement, + }, + "simple map": { + input: map[string]int{"a": 1, "b": 2}, + wantOutput: "a = 1\nb = 2\n", + }, + "map with interface{} value type": { + input: map[string]interface{}{"a": 1, "b": "c"}, + wantOutput: "a = 1\nb = \"c\"\n", + }, + "map with interface{} value type, some of which are structs": { + input: map[string]interface{}{ + "a": struct{ Int int }{2}, + "b": 1, + }, + wantOutput: "b = 1\n\n[a]\n Int = 2\n", + }, + "nested map": { + input: map[string]map[string]int{ + "a": {"b": 1}, + "c": {"d": 2}, + }, + wantOutput: "[a]\n b = 1\n\n[c]\n d = 2\n", + }, + "nested struct": { + input: struct{ Struct struct{ Int int } }{ + struct{ Int int }{1}, + }, + wantOutput: "[Struct]\n Int = 1\n", + }, + "nested struct and non-struct field": { + input: struct { + Struct struct{ Int int } + Bool bool + }{struct{ Int int }{1}, true}, + wantOutput: "Bool = true\n\n[Struct]\n Int = 1\n", + }, + "2 nested structs": { + input: struct{ Struct1, Struct2 struct{ Int int } }{ + struct{ Int int }{1}, struct{ Int int }{2}, + }, + wantOutput: "[Struct1]\n Int = 1\n\n[Struct2]\n Int = 2\n", + }, + "deeply nested structs": { + input: struct { + Struct1, Struct2 struct{ Struct3 *struct{ Int int } } + }{ + struct{ Struct3 *struct{ Int int } }{&struct{ Int int }{1}}, + struct{ Struct3 *struct{ Int int } }{nil}, + }, + wantOutput: "[Struct1]\n [Struct1.Struct3]\n Int = 1" + + "\n\n[Struct2]\n", + }, + "nested struct with nil struct elem": { + input: struct { + Struct struct{ Inner *struct{ Int int } } + }{ + struct{ Inner *struct{ Int int } }{nil}, + }, + wantOutput: "[Struct]\n", + }, + "nested struct with no fields": { + input: struct { + Struct struct{ Inner struct{} } + }{ + struct{ Inner struct{} }{struct{}{}}, + }, + wantOutput: "[Struct]\n [Struct.Inner]\n", + }, + "struct with tags": { + input: struct { + Struct struct { + Int int `toml:"_int"` + } `toml:"_struct"` + Bool bool `toml:"_bool"` + }{ + struct { + Int int `toml:"_int"` + }{1}, true, + }, + wantOutput: "_bool = true\n\n[_struct]\n _int = 1\n", + }, + "embedded struct": { + input: struct{ Embedded }{Embedded{1}}, + wantOutput: "_int = 1\n", + }, + "embedded *struct": { + input: struct{ *Embedded }{&Embedded{1}}, + wantOutput: "_int = 1\n", + }, + "nested embedded struct": { + input: struct { + Struct struct{ Embedded } `toml:"_struct"` + }{struct{ Embedded }{Embedded{1}}}, + wantOutput: "[_struct]\n _int = 1\n", + }, + "nested embedded *struct": { + input: struct { + Struct struct{ *Embedded } `toml:"_struct"` + }{struct{ *Embedded }{&Embedded{1}}}, + wantOutput: "[_struct]\n _int = 1\n", + }, + "array of tables": { + input: struct { + Structs []*struct{ Int int } `toml:"struct"` + }{ + []*struct{ Int int }{{1}, {3}}, + }, + wantOutput: "[[struct]]\n Int = 1\n\n[[struct]]\n Int = 3\n", + }, + "array of tables order": { + input: map[string]interface{}{ + "map": map[string]interface{}{ + "zero": 5, + "arr": []map[string]int{ + map[string]int{ + "friend": 5, + }, + }, + }, + }, + wantOutput: "[map]\n zero = 5\n\n [[map.arr]]\n friend = 5\n", + }, + "(error) top-level slice": { + input: []struct{ Int int }{{1}, {2}, {3}}, + wantError: errNoKey, + }, + "(error) slice of slice": { + input: struct { + Slices [][]struct{ Int int } + }{ + [][]struct{ Int int }{{{1}}, {{2}}, {{3}}}, + }, + wantError: errArrayNoTable, + }, + "(error) map no string key": { + input: map[int]string{1: ""}, + wantError: errNonString, + }, + "(error) anonymous non-struct": { + input: struct{ NonStruct }{5}, + wantError: errAnonNonStruct, + }, + "(error) empty key name": { + input: map[string]int{"": 1}, + wantError: errAnything, + }, + "(error) empty map name": { + input: map[string]interface{}{ + "": map[string]int{"v": 1}, + }, + wantError: errAnything, + }, + } + for label, test := range tests { + encodeExpected(t, label, test.input, test.wantOutput, test.wantError) + } +} + +func TestEncodeNestedTableArrays(t *testing.T) { + type song struct { + Name string `toml:"name"` + } + type album struct { + Name string `toml:"name"` + Songs []song `toml:"songs"` + } + type springsteen struct { + Albums []album `toml:"albums"` + } + value := springsteen{ + []album{ + {"Born to Run", + []song{{"Jungleland"}, {"Meeting Across the River"}}}, + {"Born in the USA", + []song{{"Glory Days"}, {"Dancing in the Dark"}}}, + }, + } + expected := `[[albums]] + name = "Born to Run" + + [[albums.songs]] + name = "Jungleland" + + [[albums.songs]] + name = "Meeting Across the River" + +[[albums]] + name = "Born in the USA" + + [[albums.songs]] + name = "Glory Days" + + [[albums.songs]] + name = "Dancing in the Dark" +` + encodeExpected(t, "nested table arrays", value, expected, nil) +} + +func TestEncodeArrayHashWithNormalHashOrder(t *testing.T) { + type Alpha struct { + V int + } + type Beta struct { + V int + } + type Conf struct { + V int + A Alpha + B []Beta + } + + val := Conf{ + V: 1, + A: Alpha{2}, + B: []Beta{{3}}, + } + expected := "V = 1\n\n[A]\n V = 2\n\n[[B]]\n V = 3\n" + encodeExpected(t, "array hash with normal hash order", val, expected, nil) +} + +func encodeExpected( + t *testing.T, label string, val interface{}, wantStr string, wantErr error, +) { + var buf bytes.Buffer + enc := NewEncoder(&buf) + err := enc.Encode(val) + if err != wantErr { + if wantErr != nil { + if wantErr == errAnything && err != nil { + return + } + t.Errorf("%s: want Encode error %v, got %v", label, wantErr, err) + } else { + t.Errorf("%s: Encode failed: %s", label, err) + } + } + if err != nil { + return + } + if got := buf.String(); wantStr != got { + t.Errorf("%s: want\n-----\n%q\n-----\nbut got\n-----\n%q\n-----\n", + label, wantStr, got) + } +} + +func ExampleEncoder_Encode() { + date, _ := time.Parse(time.RFC822, "14 Mar 10 18:00 UTC") + var config = map[string]interface{}{ + "date": date, + "counts": []int{1, 1, 2, 3, 5, 8}, + "hash": map[string]string{ + "key1": "val1", + "key2": "val2", + }, + } + buf := new(bytes.Buffer) + if err := NewEncoder(buf).Encode(config); err != nil { + log.Fatal(err) + } + fmt.Println(buf.String()) + + // Output: + // counts = [1, 1, 2, 3, 5, 8] + // date = 2010-03-14T18:00:00Z + // + // [hash] + // key1 = "val1" + // key2 = "val2" +} diff --git a/encoding_types.go b/encoding_types.go new file mode 100644 index 000000000..140c44c11 --- /dev/null +++ b/encoding_types.go @@ -0,0 +1,19 @@ +// +build go1.2 + +package toml + +// In order to support Go 1.1, we define our own TextMarshaler and +// TextUnmarshaler types. For Go 1.2+, we just alias them with the +// standard library interfaces. + +import ( + "encoding" +) + +// TextMarshaler is a synonym for encoding.TextMarshaler. It is defined here +// so that Go 1.1 can be supported. +type TextMarshaler encoding.TextMarshaler + +// TextUnmarshaler is a synonym for encoding.TextUnmarshaler. It is defined here +// so that Go 1.1 can be supported. +type TextUnmarshaler encoding.TextUnmarshaler diff --git a/encoding_types_1.1.go b/encoding_types_1.1.go new file mode 100644 index 000000000..fb285e7f5 --- /dev/null +++ b/encoding_types_1.1.go @@ -0,0 +1,18 @@ +// +build !go1.2 + +package toml + +// These interfaces were introduced in Go 1.2, so we add them manually when +// compiling for Go 1.1. + +// TextMarshaler is a synonym for encoding.TextMarshaler. It is defined here +// so that Go 1.1 can be supported. +type TextMarshaler interface { + MarshalText() (text []byte, err error) +} + +// TextUnmarshaler is a synonym for encoding.TextUnmarshaler. It is defined here +// so that Go 1.1 can be supported. +type TextUnmarshaler interface { + UnmarshalText(text []byte) error +} diff --git a/lex.go b/lex.go new file mode 100644 index 000000000..3821fa271 --- /dev/null +++ b/lex.go @@ -0,0 +1,734 @@ +package toml + +import ( + "fmt" + "strings" + "unicode/utf8" +) + +type itemType int + +const ( + itemError itemType = iota + itemNIL // used in the parser to indicate no type + itemEOF + itemText + itemString + itemBool + itemInteger + itemFloat + itemDatetime + itemArray // the start of an array + itemArrayEnd + itemTableStart + itemTableEnd + itemArrayTableStart + itemArrayTableEnd + itemKeyStart + itemCommentStart +) + +const ( + eof = 0 + tableStart = '[' + tableEnd = ']' + arrayTableStart = '[' + arrayTableEnd = ']' + tableSep = '.' + keySep = '=' + arrayStart = '[' + arrayEnd = ']' + arrayValTerm = ',' + commentStart = '#' + stringStart = '"' + stringEnd = '"' +) + +type stateFn func(lx *lexer) stateFn + +type lexer struct { + input string + start int + pos int + width int + line int + state stateFn + items chan item + + // A stack of state functions used to maintain context. + // The idea is to reuse parts of the state machine in various places. + // For example, values can appear at the top level or within arbitrarily + // nested arrays. The last state on the stack is used after a value has + // been lexed. Similarly for comments. + stack []stateFn +} + +type item struct { + typ itemType + val string + line int +} + +func (lx *lexer) nextItem() item { + for { + select { + case item := <-lx.items: + return item + default: + lx.state = lx.state(lx) + } + } +} + +func lex(input string) *lexer { + lx := &lexer{ + input: input + "\n", + state: lexTop, + line: 1, + items: make(chan item, 10), + stack: make([]stateFn, 0, 10), + } + return lx +} + +func (lx *lexer) push(state stateFn) { + lx.stack = append(lx.stack, state) +} + +func (lx *lexer) pop() stateFn { + if len(lx.stack) == 0 { + return lx.errorf("BUG in lexer: no states to pop.") + } + last := lx.stack[len(lx.stack)-1] + lx.stack = lx.stack[0 : len(lx.stack)-1] + return last +} + +func (lx *lexer) current() string { + return lx.input[lx.start:lx.pos] +} + +func (lx *lexer) emit(typ itemType) { + lx.items <- item{typ, lx.current(), lx.line} + lx.start = lx.pos +} + +func (lx *lexer) emitTrim(typ itemType) { + lx.items <- item{typ, strings.TrimSpace(lx.current()), lx.line} + lx.start = lx.pos +} + +func (lx *lexer) next() (r rune) { + if lx.pos >= len(lx.input) { + lx.width = 0 + return eof + } + + if lx.input[lx.pos] == '\n' { + lx.line++ + } + r, lx.width = utf8.DecodeRuneInString(lx.input[lx.pos:]) + lx.pos += lx.width + return r +} + +// ignore skips over the pending input before this point. +func (lx *lexer) ignore() { + lx.start = lx.pos +} + +// backup steps back one rune. Can be called only once per call of next. +func (lx *lexer) backup() { + lx.pos -= lx.width + if lx.pos < len(lx.input) && lx.input[lx.pos] == '\n' { + lx.line-- + } +} + +// accept consumes the next rune if it's equal to `valid`. +func (lx *lexer) accept(valid rune) bool { + if lx.next() == valid { + return true + } + lx.backup() + return false +} + +// peek returns but does not consume the next rune in the input. +func (lx *lexer) peek() rune { + r := lx.next() + lx.backup() + return r +} + +// errorf stops all lexing by emitting an error and returning `nil`. +// Note that any value that is a character is escaped if it's a special +// character (new lines, tabs, etc.). +func (lx *lexer) errorf(format string, values ...interface{}) stateFn { + lx.items <- item{ + itemError, + fmt.Sprintf(format, values...), + lx.line, + } + return nil +} + +// lexTop consumes elements at the top level of TOML data. +func lexTop(lx *lexer) stateFn { + r := lx.next() + if isWhitespace(r) || isNL(r) { + return lexSkip(lx, lexTop) + } + + switch r { + case commentStart: + lx.push(lexTop) + return lexCommentStart + case tableStart: + return lexTableStart + case eof: + if lx.pos > lx.start { + return lx.errorf("Unexpected EOF.") + } + lx.emit(itemEOF) + return nil + } + + // At this point, the only valid item can be a key, so we back up + // and let the key lexer do the rest. + lx.backup() + lx.push(lexTopEnd) + return lexKeyStart +} + +// lexTopEnd is entered whenever a top-level item has been consumed. (A value +// or a table.) It must see only whitespace, and will turn back to lexTop +// upon a new line. If it sees EOF, it will quit the lexer successfully. +func lexTopEnd(lx *lexer) stateFn { + r := lx.next() + switch { + case r == commentStart: + // a comment will read to a new line for us. + lx.push(lexTop) + return lexCommentStart + case isWhitespace(r): + return lexTopEnd + case isNL(r): + lx.ignore() + return lexTop + case r == eof: + lx.ignore() + return lexTop + } + return lx.errorf("Expected a top-level item to end with a new line, "+ + "comment or EOF, but got %q instead.", r) +} + +// lexTable lexes the beginning of a table. Namely, it makes sure that +// it starts with a character other than '.' and ']'. +// It assumes that '[' has already been consumed. +// It also handles the case that this is an item in an array of tables. +// e.g., '[[name]]'. +func lexTableStart(lx *lexer) stateFn { + if lx.peek() == arrayTableStart { + lx.next() + lx.emit(itemArrayTableStart) + lx.push(lexArrayTableEnd) + } else { + lx.emit(itemTableStart) + lx.push(lexTableEnd) + } + return lexTableNameStart +} + +func lexTableEnd(lx *lexer) stateFn { + lx.emit(itemTableEnd) + return lexTopEnd +} + +func lexArrayTableEnd(lx *lexer) stateFn { + if r := lx.next(); r != arrayTableEnd { + return lx.errorf("Expected end of table array name delimiter %q, "+ + "but got %q instead.", arrayTableEnd, r) + } + lx.emit(itemArrayTableEnd) + return lexTopEnd +} + +func lexTableNameStart(lx *lexer) stateFn { + switch lx.next() { + case tableEnd, eof: + return lx.errorf("Unexpected end of table. (Tables cannot " + + "be empty.)") + case tableSep: + return lx.errorf("Unexpected table separator. (Tables cannot " + + "be empty.)") + } + return lexTableName +} + +// lexTableName lexes the name of a table. It assumes that at least one +// valid character for the table has already been read. +func lexTableName(lx *lexer) stateFn { + switch lx.peek() { + case eof: + return lx.errorf("Unexpected end of table name %q.", lx.current()) + case tableStart: + return lx.errorf("Table names cannot contain %q or %q.", + tableStart, tableEnd) + case tableEnd: + lx.emit(itemText) + lx.next() + return lx.pop() + case tableSep: + lx.emit(itemText) + lx.next() + lx.ignore() + return lexTableNameStart + } + lx.next() + return lexTableName +} + +// lexKeyStart consumes a key name up until the first non-whitespace character. +// lexKeyStart will ignore whitespace. +func lexKeyStart(lx *lexer) stateFn { + r := lx.peek() + switch { + case r == keySep: + return lx.errorf("Unexpected key separator %q.", keySep) + case isWhitespace(r) || isNL(r): + lx.next() + return lexSkip(lx, lexKeyStart) + } + + lx.ignore() + lx.emit(itemKeyStart) + lx.next() + return lexKey +} + +// lexKey consumes the text of a key. Assumes that the first character (which +// is not whitespace) has already been consumed. +func lexKey(lx *lexer) stateFn { + r := lx.peek() + + // Keys cannot contain a '#' character. + if r == commentStart { + return lx.errorf("Key cannot contain a '#' character.") + } + + // XXX: Possible divergence from spec? + // "Keys start with the first non-whitespace character and end with the + // last non-whitespace character before the equals sign." + // Note here that whitespace is either a tab or a space. + // But we'll call it quits if we see a new line too. + if isNL(r) { + lx.emitTrim(itemText) + return lexKeyEnd + } + + // Let's also call it quits if we see an equals sign. + if r == keySep { + lx.emitTrim(itemText) + return lexKeyEnd + } + + lx.next() + return lexKey +} + +// lexKeyEnd consumes the end of a key (up to the key separator). +// Assumes that any whitespace after a key has been consumed. +func lexKeyEnd(lx *lexer) stateFn { + r := lx.next() + if r == keySep { + return lexSkip(lx, lexValue) + } + return lx.errorf("Expected key separator %q, but got %q instead.", + keySep, r) +} + +// lexValue starts the consumption of a value anywhere a value is expected. +// lexValue will ignore whitespace. +// After a value is lexed, the last state on the next is popped and returned. +func lexValue(lx *lexer) stateFn { + // We allow whitespace to precede a value, but NOT new lines. + // In array syntax, the array states are responsible for ignoring new lines. + r := lx.next() + if isWhitespace(r) { + return lexSkip(lx, lexValue) + } + + switch { + case r == arrayStart: + lx.ignore() + lx.emit(itemArray) + return lexArrayValue + case r == stringStart: + lx.ignore() // ignore the '"' + return lexString + case r == 't': + return lexTrue + case r == 'f': + return lexFalse + case r == '-': + return lexNumberStart + case isDigit(r): + lx.backup() // avoid an extra state and use the same as above + return lexNumberOrDateStart + case r == '.': // special error case, be kind to users + return lx.errorf("Floats must start with a digit, not '.'.") + } + return lx.errorf("Expected value but found %q instead.", r) +} + +// lexArrayValue consumes one value in an array. It assumes that '[' or ',' +// have already been consumed. All whitespace and new lines are ignored. +func lexArrayValue(lx *lexer) stateFn { + r := lx.next() + switch { + case isWhitespace(r) || isNL(r): + return lexSkip(lx, lexArrayValue) + case r == commentStart: + lx.push(lexArrayValue) + return lexCommentStart + case r == arrayValTerm: + return lx.errorf("Unexpected array value terminator %q.", + arrayValTerm) + case r == arrayEnd: + return lexArrayEnd + } + + lx.backup() + lx.push(lexArrayValueEnd) + return lexValue +} + +// lexArrayValueEnd consumes the cruft between values of an array. Namely, +// it ignores whitespace and expects either a ',' or a ']'. +func lexArrayValueEnd(lx *lexer) stateFn { + r := lx.next() + switch { + case isWhitespace(r) || isNL(r): + return lexSkip(lx, lexArrayValueEnd) + case r == commentStart: + lx.push(lexArrayValueEnd) + return lexCommentStart + case r == arrayValTerm: + lx.ignore() + return lexArrayValue // move on to the next value + case r == arrayEnd: + return lexArrayEnd + } + return lx.errorf("Expected an array value terminator %q or an array "+ + "terminator %q, but got %q instead.", arrayValTerm, arrayEnd, r) +} + +// lexArrayEnd finishes the lexing of an array. It assumes that a ']' has +// just been consumed. +func lexArrayEnd(lx *lexer) stateFn { + lx.ignore() + lx.emit(itemArrayEnd) + return lx.pop() +} + +// lexString consumes the inner contents of a string. It assumes that the +// beginning '"' has already been consumed and ignored. +func lexString(lx *lexer) stateFn { + r := lx.next() + switch { + case isNL(r): + return lx.errorf("Strings cannot contain new lines.") + case r == '\\': + return lexStringEscape + case r == stringEnd: + lx.backup() + lx.emit(itemString) + lx.next() + lx.ignore() + return lx.pop() + } + return lexString +} + +// lexStringEscape consumes an escaped character. It assumes that the preceding +// '\\' has already been consumed. +func lexStringEscape(lx *lexer) stateFn { + r := lx.next() + switch r { + case 'b': + fallthrough + case 't': + fallthrough + case 'n': + fallthrough + case 'f': + fallthrough + case 'r': + fallthrough + case '"': + fallthrough + case '/': + fallthrough + case '\\': + return lexString + case 'u': + return lexStringUnicode + } + return lx.errorf("Invalid escape character %q. Only the following "+ + "escape characters are allowed: "+ + "\\b, \\t, \\n, \\f, \\r, \\\", \\/, \\\\, and \\uXXXX.", r) +} + +// lexStringBinary consumes two hexadecimal digits following '\x'. It assumes +// that the '\x' has already been consumed. +func lexStringUnicode(lx *lexer) stateFn { + var r rune + + for i := 0; i < 4; i++ { + r = lx.next() + if !isHexadecimal(r) { + return lx.errorf("Expected four hexadecimal digits after '\\x', "+ + "but got '%s' instead.", lx.current()) + } + } + return lexString +} + +// lexNumberOrDateStart consumes either a (positive) integer, float or datetime. +// It assumes that NO negative sign has been consumed. +func lexNumberOrDateStart(lx *lexer) stateFn { + r := lx.next() + if !isDigit(r) { + if r == '.' { + return lx.errorf("Floats must start with a digit, not '.'.") + } else { + return lx.errorf("Expected a digit but got %q.", r) + } + } + return lexNumberOrDate +} + +// lexNumberOrDate consumes either a (positive) integer, float or datetime. +func lexNumberOrDate(lx *lexer) stateFn { + r := lx.next() + switch { + case r == '-': + if lx.pos-lx.start != 5 { + return lx.errorf("All ISO8601 dates must be in full Zulu form.") + } + return lexDateAfterYear + case isDigit(r): + return lexNumberOrDate + case r == '.': + return lexFloatStart + } + + lx.backup() + lx.emit(itemInteger) + return lx.pop() +} + +// lexDateAfterYear consumes a full Zulu Datetime in ISO8601 format. +// It assumes that "YYYY-" has already been consumed. +func lexDateAfterYear(lx *lexer) stateFn { + formats := []rune{ + // digits are '0'. + // everything else is direct equality. + '0', '0', '-', '0', '0', + 'T', + '0', '0', ':', '0', '0', ':', '0', '0', + 'Z', + } + for _, f := range formats { + r := lx.next() + if f == '0' { + if !isDigit(r) { + return lx.errorf("Expected digit in ISO8601 datetime, "+ + "but found %q instead.", r) + } + } else if f != r { + return lx.errorf("Expected %q in ISO8601 datetime, "+ + "but found %q instead.", f, r) + } + } + lx.emit(itemDatetime) + return lx.pop() +} + +// lexNumberStart consumes either an integer or a float. It assumes that a +// negative sign has already been read, but that *no* digits have been consumed. +// lexNumberStart will move to the appropriate integer or float states. +func lexNumberStart(lx *lexer) stateFn { + // we MUST see a digit. Even floats have to start with a digit. + r := lx.next() + if !isDigit(r) { + if r == '.' { + return lx.errorf("Floats must start with a digit, not '.'.") + } else { + return lx.errorf("Expected a digit but got %q.", r) + } + } + return lexNumber +} + +// lexNumber consumes an integer or a float after seeing the first digit. +func lexNumber(lx *lexer) stateFn { + r := lx.next() + switch { + case isDigit(r): + return lexNumber + case r == '.': + return lexFloatStart + } + + lx.backup() + lx.emit(itemInteger) + return lx.pop() +} + +// lexFloatStart starts the consumption of digits of a float after a '.'. +// Namely, at least one digit is required. +func lexFloatStart(lx *lexer) stateFn { + r := lx.next() + if !isDigit(r) { + return lx.errorf("Floats must have a digit after the '.', but got "+ + "%q instead.", r) + } + return lexFloat +} + +// lexFloat consumes the digits of a float after a '.'. +// Assumes that one digit has been consumed after a '.' already. +func lexFloat(lx *lexer) stateFn { + r := lx.next() + if isDigit(r) { + return lexFloat + } + + lx.backup() + lx.emit(itemFloat) + return lx.pop() +} + +// lexConst consumes the s[1:] in s. It assumes that s[0] has already been +// consumed. +func lexConst(lx *lexer, s string) stateFn { + for i := range s[1:] { + if r := lx.next(); r != rune(s[i+1]) { + return lx.errorf("Expected %q, but found %q instead.", s[:i+1], + s[:i]+string(r)) + } + } + return nil +} + +// lexTrue consumes the "rue" in "true". It assumes that 't' has already +// been consumed. +func lexTrue(lx *lexer) stateFn { + if fn := lexConst(lx, "true"); fn != nil { + return fn + } + lx.emit(itemBool) + return lx.pop() +} + +// lexFalse consumes the "alse" in "false". It assumes that 'f' has already +// been consumed. +func lexFalse(lx *lexer) stateFn { + if fn := lexConst(lx, "false"); fn != nil { + return fn + } + lx.emit(itemBool) + return lx.pop() +} + +// lexCommentStart begins the lexing of a comment. It will emit +// itemCommentStart and consume no characters, passing control to lexComment. +func lexCommentStart(lx *lexer) stateFn { + lx.ignore() + lx.emit(itemCommentStart) + return lexComment +} + +// lexComment lexes an entire comment. It assumes that '#' has been consumed. +// It will consume *up to* the first new line character, and pass control +// back to the last state on the stack. +func lexComment(lx *lexer) stateFn { + r := lx.peek() + if isNL(r) || r == eof { + lx.emit(itemText) + return lx.pop() + } + lx.next() + return lexComment +} + +// lexSkip ignores all slurped input and moves on to the next state. +func lexSkip(lx *lexer, nextState stateFn) stateFn { + return func(lx *lexer) stateFn { + lx.ignore() + return nextState + } +} + +// isWhitespace returns true if `r` is a whitespace character according +// to the spec. +func isWhitespace(r rune) bool { + return r == '\t' || r == ' ' +} + +func isNL(r rune) bool { + return r == '\n' || r == '\r' +} + +func isDigit(r rune) bool { + return r >= '0' && r <= '9' +} + +func isHexadecimal(r rune) bool { + return (r >= '0' && r <= '9') || + (r >= 'a' && r <= 'f') || + (r >= 'A' && r <= 'F') +} + +func (itype itemType) String() string { + switch itype { + case itemError: + return "Error" + case itemNIL: + return "NIL" + case itemEOF: + return "EOF" + case itemText: + return "Text" + case itemString: + return "String" + case itemBool: + return "Bool" + case itemInteger: + return "Integer" + case itemFloat: + return "Float" + case itemDatetime: + return "DateTime" + case itemTableStart: + return "TableStart" + case itemTableEnd: + return "TableEnd" + case itemKeyStart: + return "KeyStart" + case itemArray: + return "Array" + case itemArrayEnd: + return "ArrayEnd" + case itemCommentStart: + return "CommentStart" + } + panic(fmt.Sprintf("BUG: Unknown type '%d'.", int(itype))) +} + +func (item item) String() string { + return fmt.Sprintf("(%s, %s)", item.typ.String(), item.val) +} diff --git a/parse.go b/parse.go new file mode 100644 index 000000000..43afe3c3f --- /dev/null +++ b/parse.go @@ -0,0 +1,417 @@ +package toml + +import ( + "fmt" + "log" + "strconv" + "strings" + "time" + "unicode/utf8" +) + +type parser struct { + mapping map[string]interface{} + types map[string]tomlType + lx *lexer + + // A list of keys in the order that they appear in the TOML data. + ordered []Key + + // the full key for the current hash in scope + context Key + + // the base key name for everything except hashes + currentKey string + + // rough approximation of line number + approxLine int + + // A map of 'key.group.names' to whether they were created implicitly. + implicits map[string]bool +} + +type parseError string + +func (pe parseError) Error() string { + return string(pe) +} + +func parse(data string) (p *parser, err error) { + defer func() { + if r := recover(); r != nil { + var ok bool + if err, ok = r.(parseError); ok { + return + } + panic(r) + } + }() + + p = &parser{ + mapping: make(map[string]interface{}), + types: make(map[string]tomlType), + lx: lex(data), + ordered: make([]Key, 0), + implicits: make(map[string]bool), + } + for { + item := p.next() + if item.typ == itemEOF { + break + } + p.topLevel(item) + } + + return p, nil +} + +func (p *parser) panicf(format string, v ...interface{}) { + msg := fmt.Sprintf("Near line %d, key '%s': %s", + p.approxLine, p.current(), fmt.Sprintf(format, v...)) + panic(parseError(msg)) +} + +func (p *parser) next() item { + it := p.lx.nextItem() + if it.typ == itemError { + p.panicf("Near line %d: %s", it.line, it.val) + } + return it +} + +func (p *parser) bug(format string, v ...interface{}) { + log.Fatalf("BUG: %s\n\n", fmt.Sprintf(format, v...)) +} + +func (p *parser) expect(typ itemType) item { + it := p.next() + p.assertEqual(typ, it.typ) + return it +} + +func (p *parser) assertEqual(expected, got itemType) { + if expected != got { + p.bug("Expected '%s' but got '%s'.", expected, got) + } +} + +func (p *parser) topLevel(item item) { + switch item.typ { + case itemCommentStart: + p.approxLine = item.line + p.expect(itemText) + case itemTableStart: + kg := p.expect(itemText) + p.approxLine = kg.line + + key := make(Key, 0) + for ; kg.typ == itemText; kg = p.next() { + key = append(key, kg.val) + } + p.assertEqual(itemTableEnd, kg.typ) + + p.establishContext(key, false) + p.setType("", tomlHash) + p.ordered = append(p.ordered, key) + case itemArrayTableStart: + kg := p.expect(itemText) + p.approxLine = kg.line + + key := make(Key, 0) + for ; kg.typ == itemText; kg = p.next() { + key = append(key, kg.val) + } + p.assertEqual(itemArrayTableEnd, kg.typ) + + p.establishContext(key, true) + p.setType("", tomlArrayHash) + p.ordered = append(p.ordered, key) + case itemKeyStart: + kname := p.expect(itemText) + p.currentKey = kname.val + p.approxLine = kname.line + + val, typ := p.value(p.next()) + p.setValue(p.currentKey, val) + p.setType(p.currentKey, typ) + p.ordered = append(p.ordered, p.context.add(p.currentKey)) + + p.currentKey = "" + default: + p.bug("Unexpected type at top level: %s", item.typ) + } +} + +// value translates an expected value from the lexer into a Go value wrapped +// as an empty interface. +func (p *parser) value(it item) (interface{}, tomlType) { + switch it.typ { + case itemString: + return p.replaceUnicode(replaceEscapes(it.val)), p.typeOfPrimitive(it) + case itemBool: + switch it.val { + case "true": + return true, p.typeOfPrimitive(it) + case "false": + return false, p.typeOfPrimitive(it) + } + p.bug("Expected boolean value, but got '%s'.", it.val) + case itemInteger: + num, err := strconv.ParseInt(it.val, 10, 64) + if err != nil { + // See comment below for floats describing why we make a + // distinction between a bug and a user error. + if e, ok := err.(*strconv.NumError); ok && + e.Err == strconv.ErrRange { + + p.panicf("Integer '%s' is out of the range of 64-bit "+ + "signed integers.", it.val) + } else { + p.bug("Expected integer value, but got '%s'.", it.val) + } + } + return num, p.typeOfPrimitive(it) + case itemFloat: + num, err := strconv.ParseFloat(it.val, 64) + if err != nil { + // Distinguish float values. Normally, it'd be a bug if the lexer + // provides an invalid float, but it's possible that the float is + // out of range of valid values (which the lexer cannot determine). + // So mark the former as a bug but the latter as a legitimate user + // error. + // + // This is also true for integers. + if e, ok := err.(*strconv.NumError); ok && + e.Err == strconv.ErrRange { + + p.panicf("Float '%s' is out of the range of 64-bit "+ + "IEEE-754 floating-point numbers.", it.val) + } else { + p.bug("Expected float value, but got '%s'.", it.val) + } + } + return num, p.typeOfPrimitive(it) + case itemDatetime: + t, err := time.Parse("2006-01-02T15:04:05Z", it.val) + if err != nil { + p.bug("Expected Zulu formatted DateTime, but got '%s'.", it.val) + } + return t, p.typeOfPrimitive(it) + case itemArray: + array := make([]interface{}, 0) + types := make([]tomlType, 0) + + for it = p.next(); it.typ != itemArrayEnd; it = p.next() { + if it.typ == itemCommentStart { + p.expect(itemText) + continue + } + + val, typ := p.value(it) + array = append(array, val) + types = append(types, typ) + } + return array, p.typeOfArray(types) + } + p.bug("Unexpected value type: %s", it.typ) + panic("unreachable") +} + +// establishContext sets the current context of the parser, +// where the context is either a hash or an array of hashes. Which one is +// set depends on the value of the `array` parameter. +// +// Establishing the context also makes sure that the key isn't a duplicate, and +// will create implicit hashes automatically. +func (p *parser) establishContext(key Key, array bool) { + var ok bool + + // Always start at the top level and drill down for our context. + hashContext := p.mapping + keyContext := make(Key, 0) + + // We only need implicit hashes for key[0:-1] + for _, k := range key[0 : len(key)-1] { + _, ok = hashContext[k] + keyContext = append(keyContext, k) + + // No key? Make an implicit hash and move on. + if !ok { + p.addImplicit(keyContext) + hashContext[k] = make(map[string]interface{}) + } + + // If the hash context is actually an array of tables, then set + // the hash context to the last element in that array. + // + // Otherwise, it better be a table, since this MUST be a key group (by + // virtue of it not being the last element in a key). + switch t := hashContext[k].(type) { + case []map[string]interface{}: + hashContext = t[len(t)-1] + case map[string]interface{}: + hashContext = t + default: + p.panicf("Key '%s' was already created as a hash.", keyContext) + } + } + + p.context = keyContext + if array { + // If this is the first element for this array, then allocate a new + // list of tables for it. + k := key[len(key)-1] + if _, ok := hashContext[k]; !ok { + hashContext[k] = make([]map[string]interface{}, 0, 5) + } + + // Add a new table. But make sure the key hasn't already been used + // for something else. + if hash, ok := hashContext[k].([]map[string]interface{}); ok { + hashContext[k] = append(hash, make(map[string]interface{})) + } else { + p.panicf("Key '%s' was already created and cannot be used as "+ + "an array.", keyContext) + } + } else { + p.setValue(key[len(key)-1], make(map[string]interface{})) + } + p.context = append(p.context, key[len(key)-1]) +} + +// setValue sets the given key to the given value in the current context. +// It will make sure that the key hasn't already been defined, account for +// implicit key groups. +func (p *parser) setValue(key string, value interface{}) { + var tmpHash interface{} + var ok bool + + hash := p.mapping + keyContext := make(Key, 0) + for _, k := range p.context { + keyContext = append(keyContext, k) + if tmpHash, ok = hash[k]; !ok { + p.bug("Context for key '%s' has not been established.", keyContext) + } + switch t := tmpHash.(type) { + case []map[string]interface{}: + // The context is a table of hashes. Pick the most recent table + // defined as the current hash. + hash = t[len(t)-1] + case map[string]interface{}: + hash = t + default: + p.bug("Expected hash to have type 'map[string]interface{}', but "+ + "it has '%T' instead.", tmpHash) + } + } + keyContext = append(keyContext, key) + + if _, ok := hash[key]; ok { + // Typically, if the given key has already been set, then we have + // to raise an error since duplicate keys are disallowed. However, + // it's possible that a key was previously defined implicitly. In this + // case, it is allowed to be redefined concretely. (See the + // `tests/valid/implicit-and-explicit-after.toml` test in `toml-test`.) + // + // But we have to make sure to stop marking it as an implicit. (So that + // another redefinition provokes an error.) + // + // Note that since it has already been defined (as a hash), we don't + // want to overwrite it. So our business is done. + if p.isImplicit(keyContext) { + p.removeImplicit(keyContext) + return + } + + // Otherwise, we have a concrete key trying to override a previous + // key, which is *always* wrong. + p.panicf("Key '%s' has already been defined.", keyContext) + } + hash[key] = value +} + +// setType sets the type of a particular value at a given key. +// It should be called immediately AFTER setValue. +// +// Note that if `key` is empty, then the type given will be applied to the +// current context (which is either a table or an array of tables). +func (p *parser) setType(key string, typ tomlType) { + keyContext := make(Key, 0, len(p.context)+1) + for _, k := range p.context { + keyContext = append(keyContext, k) + } + if len(key) > 0 { // allow type setting for hashes + keyContext = append(keyContext, key) + } + p.types[keyContext.String()] = typ +} + +// addImplicit sets the given Key as having been created implicitly. +func (p *parser) addImplicit(key Key) { + p.implicits[key.String()] = true +} + +// removeImplicit stops tagging the given key as having been implicitly created. +func (p *parser) removeImplicit(key Key) { + p.implicits[key.String()] = false +} + +// isImplicit returns true if the key group pointed to by the key was created +// implicitly. +func (p *parser) isImplicit(key Key) bool { + return p.implicits[key.String()] +} + +// current returns the full key name of the current context. +func (p *parser) current() string { + if len(p.currentKey) == 0 { + return p.context.String() + } + if len(p.context) == 0 { + return p.currentKey + } + return fmt.Sprintf("%s.%s", p.context, p.currentKey) +} + +func replaceEscapes(s string) string { + return strings.NewReplacer( + "\\b", "\u0008", + "\\t", "\u0009", + "\\n", "\u000A", + "\\f", "\u000C", + "\\r", "\u000D", + "\\\"", "\u0022", + "\\/", "\u002F", + "\\\\", "\u005C", + ).Replace(s) +} + +func (p *parser) replaceUnicode(s string) string { + indexEsc := func() int { + return strings.Index(s, "\\u") + } + for i := indexEsc(); i != -1; i = indexEsc() { + asciiBytes := s[i+2 : i+6] + s = strings.Replace(s, s[i:i+6], p.asciiEscapeToUnicode(asciiBytes), -1) + } + return s +} + +func (p *parser) asciiEscapeToUnicode(s string) string { + hex, err := strconv.ParseUint(strings.ToLower(s), 16, 32) + if err != nil { + p.bug("Could not parse '%s' as a hexadecimal number, but the "+ + "lexer claims it's OK: %s", s, err) + } + + // BUG(burntsushi) + // I honestly don't understand how this works. I can't seem + // to find a way to make this fail. I figured this would fail on invalid + // UTF-8 characters like U+DCFF, but it doesn't. + r := string(rune(hex)) + if !utf8.ValidString(r) { + p.panicf("Escaped character '\\u%s' is not valid UTF-8.", s) + } + return string(r) +} diff --git a/session.vim b/session.vim new file mode 100644 index 000000000..562164be0 --- /dev/null +++ b/session.vim @@ -0,0 +1 @@ +au BufWritePost *.go silent!make tags > /dev/null 2>&1 diff --git a/type_check.go b/type_check.go new file mode 100644 index 000000000..79dac6b19 --- /dev/null +++ b/type_check.go @@ -0,0 +1,85 @@ +package toml + +// tomlType represents any Go type that corresponds to a TOML type. +// While the first draft of the TOML spec has a simplistic type system that +// probably doesn't need this level of sophistication, we seem to be militating +// toward adding real composite types. +type tomlType interface { + typeString() string +} + +// typeEqual accepts any two types and returns true if they are equal. +func typeEqual(t1, t2 tomlType) bool { + if t1 == nil || t2 == nil { + return false + } + return t1.typeString() == t2.typeString() +} + +func typeIsHash(t tomlType) bool { + return typeEqual(t, tomlHash) || typeEqual(t, tomlArrayHash) +} + +type tomlBaseType string + +func (btype tomlBaseType) typeString() string { + return string(btype) +} + +func (btype tomlBaseType) String() string { + return btype.typeString() +} + +var ( + tomlInteger tomlBaseType = "Integer" + tomlFloat tomlBaseType = "Float" + tomlDatetime tomlBaseType = "Datetime" + tomlString tomlBaseType = "String" + tomlBool tomlBaseType = "Bool" + tomlArray tomlBaseType = "Array" + tomlHash tomlBaseType = "Hash" + tomlArrayHash tomlBaseType = "ArrayHash" +) + +// typeOfPrimitive returns a tomlType of any primitive value in TOML. +// Primitive values are: Integer, Float, Datetime, String and Bool. +// +// Passing a lexer item other than the following will cause a BUG message +// to occur: itemString, itemBool, itemInteger, itemFloat, itemDatetime. +func (p *parser) typeOfPrimitive(lexItem item) tomlType { + switch lexItem.typ { + case itemInteger: + return tomlInteger + case itemFloat: + return tomlFloat + case itemDatetime: + return tomlDatetime + case itemString: + return tomlString + case itemBool: + return tomlBool + } + p.bug("Cannot infer primitive type of lex item '%s'.", lexItem) + panic("unreachable") +} + +// typeOfArray returns a tomlType for an array given a list of types of its +// values. +// +// In the current spec, if an array is homogeneous, then its type is always +// "Array". If the array is not homogeneous, an error is generated. +func (p *parser) typeOfArray(types []tomlType) tomlType { + // Empty arrays are cool. + if len(types) == 0 { + return tomlArray + } + + theType := types[0] + for _, t := range types[1:] { + if !typeEqual(theType, t) { + p.panicf("Array contains values of type '%s' and '%s', but arrays "+ + "must be homogeneous.", theType, t) + } + } + return tomlArray +} diff --git a/type_fields.go b/type_fields.go new file mode 100644 index 000000000..7592f87a4 --- /dev/null +++ b/type_fields.go @@ -0,0 +1,241 @@ +package toml + +// Struct field handling is adapted from code in encoding/json: +// +// Copyright 2010 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the Go distribution. + +import ( + "reflect" + "sort" + "sync" +) + +// A field represents a single field found in a struct. +type field struct { + name string // the name of the field (`toml` tag included) + tag bool // whether field has a `toml` tag + index []int // represents the depth of an anonymous field + typ reflect.Type // the type of the field +} + +// byName sorts field by name, breaking ties with depth, +// then breaking ties with "name came from toml tag", then +// breaking ties with index sequence. +type byName []field + +func (x byName) Len() int { return len(x) } + +func (x byName) Swap(i, j int) { x[i], x[j] = x[j], x[i] } + +func (x byName) Less(i, j int) bool { + if x[i].name != x[j].name { + return x[i].name < x[j].name + } + if len(x[i].index) != len(x[j].index) { + return len(x[i].index) < len(x[j].index) + } + if x[i].tag != x[j].tag { + return x[i].tag + } + return byIndex(x).Less(i, j) +} + +// byIndex sorts field by index sequence. +type byIndex []field + +func (x byIndex) Len() int { return len(x) } + +func (x byIndex) Swap(i, j int) { x[i], x[j] = x[j], x[i] } + +func (x byIndex) Less(i, j int) bool { + for k, xik := range x[i].index { + if k >= len(x[j].index) { + return false + } + if xik != x[j].index[k] { + return xik < x[j].index[k] + } + } + return len(x[i].index) < len(x[j].index) +} + +// typeFields returns a list of fields that TOML should recognize for the given +// type. The algorithm is breadth-first search over the set of structs to +// include - the top struct and then any reachable anonymous structs. +func typeFields(t reflect.Type) []field { + // Anonymous fields to explore at the current level and the next. + current := []field{} + next := []field{{typ: t}} + + // Count of queued names for current level and the next. + count := map[reflect.Type]int{} + nextCount := map[reflect.Type]int{} + + // Types already visited at an earlier level. + visited := map[reflect.Type]bool{} + + // Fields found. + var fields []field + + for len(next) > 0 { + current, next = next, current[:0] + count, nextCount = nextCount, map[reflect.Type]int{} + + for _, f := range current { + if visited[f.typ] { + continue + } + visited[f.typ] = true + + // Scan f.typ for fields to include. + for i := 0; i < f.typ.NumField(); i++ { + sf := f.typ.Field(i) + if sf.PkgPath != "" { // unexported + continue + } + name := sf.Tag.Get("toml") + if name == "-" { + continue + } + index := make([]int, len(f.index)+1) + copy(index, f.index) + index[len(f.index)] = i + + ft := sf.Type + if ft.Name() == "" && ft.Kind() == reflect.Ptr { + // Follow pointer. + ft = ft.Elem() + } + + // Record found field and index sequence. + if name != "" || !sf.Anonymous || ft.Kind() != reflect.Struct { + tagged := name != "" + if name == "" { + name = sf.Name + } + fields = append(fields, field{name, tagged, index, ft}) + if count[f.typ] > 1 { + // If there were multiple instances, add a second, + // so that the annihilation code will see a duplicate. + // It only cares about the distinction between 1 or 2, + // so don't bother generating any more copies. + fields = append(fields, fields[len(fields)-1]) + } + continue + } + + // Record new anonymous struct to explore in next round. + nextCount[ft]++ + if nextCount[ft] == 1 { + f := field{name: ft.Name(), index: index, typ: ft} + next = append(next, f) + } + } + } + } + + sort.Sort(byName(fields)) + + // Delete all fields that are hidden by the Go rules for embedded fields, + // except that fields with TOML tags are promoted. + + // The fields are sorted in primary order of name, secondary order + // of field index length. Loop over names; for each name, delete + // hidden fields by choosing the one dominant field that survives. + out := fields[:0] + for advance, i := 0, 0; i < len(fields); i += advance { + // One iteration per name. + // Find the sequence of fields with the name of this first field. + fi := fields[i] + name := fi.name + for advance = 1; i+advance < len(fields); advance++ { + fj := fields[i+advance] + if fj.name != name { + break + } + } + if advance == 1 { // Only one field with this name + out = append(out, fi) + continue + } + dominant, ok := dominantField(fields[i : i+advance]) + if ok { + out = append(out, dominant) + } + } + + fields = out + sort.Sort(byIndex(fields)) + + return fields +} + +// dominantField looks through the fields, all of which are known to +// have the same name, to find the single field that dominates the +// others using Go's embedding rules, modified by the presence of +// TOML tags. If there are multiple top-level fields, the boolean +// will be false: This condition is an error in Go and we skip all +// the fields. +func dominantField(fields []field) (field, bool) { + // The fields are sorted in increasing index-length order. The winner + // must therefore be one with the shortest index length. Drop all + // longer entries, which is easy: just truncate the slice. + length := len(fields[0].index) + tagged := -1 // Index of first tagged field. + for i, f := range fields { + if len(f.index) > length { + fields = fields[:i] + break + } + if f.tag { + if tagged >= 0 { + // Multiple tagged fields at the same level: conflict. + // Return no field. + return field{}, false + } + tagged = i + } + } + if tagged >= 0 { + return fields[tagged], true + } + // All remaining fields have the same length. If there's more than one, + // we have a conflict (two fields named "X" at the same level) and we + // return no field. + if len(fields) > 1 { + return field{}, false + } + return fields[0], true +} + +var fieldCache struct { + sync.RWMutex + m map[reflect.Type][]field +} + +// cachedTypeFields is like typeFields but uses a cache to avoid repeated work. +func cachedTypeFields(t reflect.Type) []field { + fieldCache.RLock() + f := fieldCache.m[t] + fieldCache.RUnlock() + if f != nil { + return f + } + + // Compute fields without lock. + // Might duplicate effort but won't hold other computations back. + f = typeFields(t) + if f == nil { + f = []field{} + } + + fieldCache.Lock() + if fieldCache.m == nil { + fieldCache.m = map[reflect.Type][]field{} + } + fieldCache.m[t] = f + fieldCache.Unlock() + return f +}