- ...
- Update to libpg_query 16-5.1.0
- Add support for running on Windows
- Add support for compiling on 32-bit systems
- Treewalker: Allow passing a block with a single argument to walk!
- PgQuery::Node: Add inner and inner= helpers to get/set inner object
- Upgrade to libpg_query 16-5.0.0
- Updates to the Postgres 16 parser
- Multiple deparser improvements
- Update to libpg_query 15-4.2.3
- Update to libpg_query 15-4.2.2
- Deparser: Add support for multi-statement CREATE PROCEDURE definitions
- Deparser: Correctly quote identifier in ALTER TABLE ... ADD CONSTRAINT [x]
- Deparser: Add support for index fillfactor within CREATE TABLE, fix SHOW ALL
- Fix builds on FreeBSD (#292)
- This was broken since 4.2.0, due to pg_query_ruby_freebsd.sym being removed by accident
- Parse: Fix
ALTER INDEX my_index_name
to returntables=[]
(#285) - Parse: Detect tables in a SELECT INTO clause as DDL tables (#281)
- Add support for Ruby 3.2 (#283)
- Bump up
google-protobuf
dependency to>= 3.22.3
- 3.22.0 or newer is required for Ruby 3.2 support
- Update to libpg_query 15-4.2.1
- Deparser: Handle INTERVAL correctly when used in SET statements
- Deparser: Ensure index names are quoted as identifiers
- Update to libpg_query 15-4.2.0
- Update to PostgreSQL 15.1
- Detect tables used in the query of a PREPARE statement (#273)
- Expose recursive walk functionality via walk! (#268)
- Retain schema in name when parsing out functions (#272)
- Update to libpg_query 13-2.2.0 (#264)
- Fingerprinting version 3.1
- Fixes issue with "SELECT DISTINCT" having the same fingerprint as "SELECT" (fingerprints for "SELECT DISTINCT" will change with this revision)
- Group additional DDL statements together that otherwise generate a lot of unique fingerprints (ListenStmt, UnlistenStmt, NotifyStmt, CreateFunctionStmt, FunctionParameter and DoStmt)
- Deparser improvements
- Prefix errors with "deparse", and remove some asserts
- Fix potential segfault when passing invalid protobuf (RawStmt without Stmt)
- Update to Postgres 13.8 patch release
- Backport Xcode 14.1 build fix from upcoming 13.9 release
- Normalize additional DDL statements
- Add support for analyzing PL/pgSQL code inside DO blocks
- Fix memory leak in pg_query_fingerprint error handling
- PL/pgSQL parser: Add support for Assert, SET, COMMIT, ROLLBACK and CALL
- Add support for parsing more operators that include a
?
character
- Fingerprinting version 3.1
- Support deparsing deeply nested queries (#259)
- Truncate: Simplify VALUES(...) lists
- Truncate: Correctly handle UPDATE and ON CONFLICT target lists
- Support complex queries with deeply nested ASTs (#238)
- Find table references inside type casts
- Find function calls referenced in expression indexes (#249)
- Drop
Init_pg_query
from exported symbol map (#256)
- Track tables in EXCEPT and INTERSECT queries (#239)
- Get filter_columns working with UNION/EXCEPT/INTERSECT (#240)
- Update google-protobuf to address CVE scanner complaints
- Note that none of the CVEs apply to pg_query, but this avoids unnecessary errors when the google-protobuf dependency is pulled in
- Find tables in using clause of delete statement (#234)
- Find tables in case statements (#235)
- Correctly find nested tables in a subselect in a join condition (#233)
- Mark Postgres methods as visibility hidden, to avoid bloating dynamic symbol table (#232)
- This is required on ELF platforms (i.e. Linux, etc) to avoid including all global symbols in the shared library's symbol table, bloating the size, and causing potential conflicts with other C libraries using the same symbol names.
- Update to libpg_query 13-2.1.0 (#230)
- Normalize: add funcname error object
- Normalize: Match GROUP BY against target list and re-use param refs
- PL/pgSQL: Setup namespace items for parameters, support RECORD types
- This significantly improves parsing for PL/pgSQL functions, to the extent that most functions should now parse successfully
- Normalize: Don't modify constants in TypeName typmods/arrayBounds fields
- This matches how pg_stat_statement behaves, and avoids causing parsing errors on the normalized statement
- Don't fail builds on systems that have strchrnul support (FreeBSD)
- Fix build on FreeBSD (#222)
- Add workaround for Ruby garbage collection bug (#227)
- The Ruby interpreter has a bug in
String#concat
where the appended array may be garbage collected prematurely because the compiler optimized out a Ruby stack variable. We now callto_ary
on the Protobuf object to ensure the array lands on the Ruby stack so the garbage collector sees it. - The real fix in the interpreter is described in https://bugs.ruby-lang.org/issues/18140#note-2, but most current Ruby interpreters won't have this fix for some time.
- The Ruby interpreter has a bug in
- Table/function extraction: Support subselects and LATERAL better (#229)
- This reworks the parsing logic so we don't ignore certain kinds of subselects.
- Update to libpg_query 13-2.0.6
- Update to Postgres 13.3 patch release
- Normalize: Don't touch "GROUP BY 1" and "ORDER BY 1" expressions, keep original text
- Fingerprint: Cache list item hashes to fingerprint complex queries faster
- Deparser: Emit the RangeVar catalogname if present
- Fix crash in pg_scan function when encountering backslash escapes
- Support extracting functions from a parsed query (#147)
- Adds new
functions
,ddl_functions
andcall_functions
methods - Note that functions are identified by their name only, not their full type definition, since raw query parsetrees don't contain sufficient data to identify the types of arguments when functions are called
- Adds new
- Relax google-protobuf dependency (#213)
- Update google-protobuf to 3.17.1 (#212)
- google-protobuf 3.15.x has a bug that causes a seg fault in Ruby under certain conditions (protocolbuffers/protobuf#8639). Use google-protobuf 3.17.1 instead.
- Use Protobuf definition for determining JSON field names
- Note you may see a breaking change if you were using
PgQuery::ParseResult.encode_json
to map the protobuf result to JSON, since this now respects the intended JSON names from the Proto3 definition (instead of the differently formatted Protobuf field names)
- Note you may see a breaking change if you were using
- Rakefile: Fix "rake clean" by using CLEAN.include instead of CLEAN.<<
- Find tables inside COALESCE/MIN/MAX functions, UPDATE FROM list
- Extconf: Add library include path using $INCFLAGS, list it first
- This ensures any system installed libpg_query gets considered after the bundled libpg_query, avoiding errors where the wrong header files are used.
- Update to libpg_query 13-2.0.4
- Normalize: Fix handling of two subsequent DefElem elements (avoids crash)
- Deparser: Fix crash in CopyStmt with HEADER or FREEZE inside WITH parens
COALESCE
arguments are now included in#filter_columns
- Improve error message for protobuf parse failures
- Extconf: Fix object path regexp
- This accidentally replaced
.c
in the wrong parts of the path in some cases, causing build failures
- This accidentally replaced
- Update to libpg_query 13-2.0.2
- Fix ARM builds: Avoid dependency on cpuid.h header
- Simplify deparser of TableLikeClause
- Fix asprintf warnings by ensuring _GNU_SOURCE is set early enough
- Fix gemspec to correctly reference include files
- This would have shown as a build failure when using the published
2.0.0
gem
- This would have shown as a build failure when using the published
- Update to PostgreSQL 13 parser
- Update to libpg_query v2, and new Protobuf-based format
- WARNING: This is a breaking change if you are directly interacting with the
parsetree (helpers like
table
and such still work the same way)
- WARNING: This is a breaking change if you are directly interacting with the
parsetree (helpers like
- Use actual Ruby classes for parser result, instead of decoded JSON
- This is essentialy to enable easy and fast two-way communication with the C library, and as a bonus makes for a better interaction on the Ruby side, as we are handling actual objects instead of hashes and arrays.
- Use new deparser maintained directly in libpg_query
- This replaces the complete Ruby deparser with a new, more complete deparser that is directly maintained in libpg_query. Further deparser improvements should be directly contributed to libpg_query
- Tables helper: Return more details through
#tables_with_details
method- This is renamed from the previously badly named
#tables_with_types
method. Note that this change should not affect the output of the primarytables
helper.
- This is renamed from the previously badly named
- Replace on-demand libpg_query source download with bundled source code
- Its unnecessary to download the source on-demand, and makes this more complex than it needs to be. Instead, introduce a new "update_source" rake task that can be called to refresh the source for a specified revision.
- Re-implement smart truncation without requiring a special node type
- This ensures the
#truncate
method works with the new deparser, without the C level code needing to know about it. We may add it in the C library in the future for edge cases that can't be covered by this slightly hack-ish approach, but for now this avoids unnecessary C library deparser modifications with non-standard node types.
- This ensures the
- Update Ruby fingerprinting to new fingerprint format and XXH3 hash
- Note that its recommended to use
PgQuery.fingerprint
for performance reasons, but when the tree has been modified, it can be convenient to run a Ruby-side fingerprint instead of the C-based one that is faster.
- Note that its recommended to use
- Incorporate newer libpg_query updates in 10-1.0.3 and 10-1.0.4
- Adds support for running on ARM
- Fixes an asprintf warning during builds
- Updates to newer Postgres 10 patch release (10.15)
- Deparsing improvements by @emin100
- Add support for additional DROP statements (#147)
- Fix
CREATE TABLE AS
- Support withoutTEMP
, AddON COMMIT
(#149) - Empty target list support (#156)
UNION
parentheses (#158)OVERLAY
keyword function (#161)- Array indirection (#162)
ARRAY
functions (#163)- Correctly handle column names that need escaping in
INSERT
andUPDATE
statements (#164) INSERT INTO ON CONFLICT
(#166)LATERAL JOIN
(#168)UPDATE FROM
clause (#170)SELECT
aggregateFILTER
(#175)INTERSECT
operator (#176)
- Deparsing: Improve handling of boolean type casts @himanshu-pro & @emin100
tables
method: Find tables in the subquery ofCREATE TABLE AS
(#172) @Tassosb- Support Ruby 3.0, verify SHA256 checksum of downloaded libpg_query (#178) @stanhu
- Verify SHA256 checksum to guard against any malicious attempts to change the archive
- Use
URI.open
to fix Ruby 3.0 support
- Reduce escaped keywords to Postgres-specific keywords, and ignore unreserved keywords
- This matches the behaviour of Postgres' quote_identifier function, and avoids problems when doing text comparisons with output involving that function
- Note that this will lead to different output than in earlier pg_query versions, in some cases
- Deparsing improvements by @emin100
- Deparse
ILIKE
,COLLATE
andDISCARD
(#133) CREATE CAST
(#136)CREATE SCHEMA
(#136)UNION
,UNION ALL
andEXCEPT
inSELECT
queries (#136)CREATE DOMAIN
(#145)- Subquery indirection (#157)
- Fix Type Cast Parentheses Problem (#152)
SELECT INTO
(#151)SET DEFAULT
inINSERT INTO
(#154)REVOKE
(#155)PREPARE
andEXECUTE
(#148)INSERT INTO ... RETURNING
(#153)- Fix Alter ..
RENAME SQL
(#146)
- Deparse
- Deparsing improvements by @herwinw
- Deparsing improvements by @akiellor
- Named window functions (#150)
- Deparsing improvements by @himanshu
- Use explicit -z flag to support OpenBSD tar (#134) @sirn
- Add Ruby 2.6 to Travis tests
- Escape identifiers in more cases, if necessary
- Deparsing improvements by @herwinw
- Ignore pg_query.so in git checkout #110 @herwinw
- Prefer
__dir__
overFile.dirname(__FILE__)
#110 @herwinw
- Deparsing improvements
SELECT DISTINCT
clause #77 @Papierkorb- "
CASE expr WHEN ... END
" clause #78 @Papierkorb LEFT
/RIGHT
/FULL
/NATURAL JOIN
#79 @PapierkorbSELECT
that includes schema name #80 @jcsjcs
- Parse CTEs and nested selects in INSERT/UPDATE #76 @jcoleman
- Drop explicit json dependency #74 @yuki24
- IMPORTANT: Major version bump to indicate backwards incompatible parse tree change!
- Update to Postgres 10 parser and fingerprint version 2
- This is a backwards-incompatible change in parser output format, although it should be relatively easy to update most programs. This can't be avoided since Postgres does not guarantee parse trees stay the same across versions
- Update to libpg_query 9.5-1.7.1
- Allow "
$1 FROM $2
" to be parsed (new with pg_stat_statements in Postgres 10)
- Allow "
- Update to libpg_query 9.5-1.7.0
- Fixes compilation old gcc before 4.6.0 #73
- Fix table detection for SELECTs that have sub-SELECTs without
FROM
clause #69
- Support table detection in sub-SELECTs in
JOIN
s #68 @seanmdick - Legacy ".parsetree" helper: Fix "Between" and "In" operator does not have "AEXPR" #66
- For new applications please use ".tree" method which uses the native structure returned from libpg_query which resembles Postgres node names more closely
- Fix regression in 0.13.1 that broke ".tables" logic for
COPY
statements that don't have a target table (i.e. are reading out data vs copying in)
- Introduce split between SELECT/DML/DDL for tables method #65 @chrisfrommann
- Backwards compatible, use the new select_tables/dml_tables/ddl_tables to access the categorized table references
- Update libpg_query to 9.5-1.6.2
- Update to Fingerprinting Version 1.3
- Attributes to be ignored:
- RangeVar.relname (if node also has RangeVar.relpersistence = "t")
- Special cases: List nodes where parent field name is valuesLists
- Follow same logic described for fromClause/targetList/cols/rexpr
- Attributes to be ignored:
- Update to Fingerprinting Version 1.3
- Update libpg_query to 9.5-1.6.1
- Update to Fingerprinting Version 1.2
- Ignore portalname in DeclareCursorStmt, FetchStmt and ClosePortalStmt
- Update to Fingerprinting Version 1.2
- Update libpg_query to 9.5-1.6.0
- BREAKING CHANGE in PgQuery.normalize(..) output
- This matches the change in the upcoming Postgres 10, and makes it easier to migrate applications to the new normalization format using $1..$N instead of ?
- Deparse coldeflist #64 @jcsjcs
- Use Integer class for checking integer instead of Fixnum #62 @makimoto
- Compatibility with Ruby 2.4 #59 @merqlove
- Deparse varchar and numeric casts without arguments #61 @jcsjcs
- Update to newest libpg_query version (9.5-1.4.2)
- Cut off fingerprints at 100 nodes deep to avoid excessive runtimes/memory
- Fix warning on Linux due to missing asprintf include
- Improved deparsing @jcsjcs
- Fix build on BSD systems #58 @myfreeweb
- Update to newest libpg_query version (9.5-1.4.1)
- This release makes sure we work correctly in threaded environments
- Updated fingerprinting logic to version 1.1
- Fixes an issue with UpdateStmt target lists being ignored
- Update to newest libpg_query version (9.5-1.4.0)
- Improved table name analysis (
#tables
method)
- Based on PostgreSQL 9.5.3
- Use LLVM extracted parser for significantly improved build times (via libpg_query)
- Deparsing Improvements
- Fix issue with A_CONST string values in
.parsetree
compatibility layer (Fixes #47)
- Add support for Ruby 1.9 (Fixes #44)
- Based on PostgreSQL 9.5.2
- NOTE: Output format for the parse tree has changed (backwards incompatible!),
it is recommended you extensively test any direct reading/modification of
the tree data in your own code
- You can use the
.parsetree
translator method to ease the transition, note however that there are still a few incompatible changes
- You can use the
- New
.fingerprint
method (backwards incompatible as well), see https://github.com/lfittl/libpg_query/wiki/Fingerprinting - Removes PostgreSQL source and tarball after build process has finished, to reduce diskspace requirements of the installed gem
- Use fixed git version for libpg_query (PostgreSQL 9.4 based)
- NOTE: 0.8 will be the last series with the initial parse tree format, 0.9 will introduce a newer, more stable, but backwards incompatible parse tree format
- Deparsing
- Abstracted parser access into libpg_query #24
- libpg_query
- Restructure build process to use upstream tarballs #35
- Avoid bison/flex dependency to make deployment easier #31
- Solve issues with deployments to Heroku #32
- Deparsing
HAVING
andFOR UPDATE
#36 @JackDanger
- Deparsing
- Constraints & Interval Types #28 @JackDanger
- Cross joins #29 @mme
ALTER TABLE
#30 @JackDangerLIMIT and OFFSET
#33 @jcsjcs
- Deparsing
COUNT(*)
@JackDanger- Window clauses Chris Martin
CREATE TABLE
/VIEW/FUNCTION
@JackDanger
- Return exact location for parser errors @JackDanger
- Speed up gem install by not generating rdoc/ri for the Postgres source
- Deparsing: Support
WITH
clauses inINSERT
/UPDATE
/DELETE
@JackDanger - Make sure gemspec includes all necessary files
- Deparsing (experimental)
- Turns parse trees into SQL again
- New truncate method to smartly truncate based on less important query parts
- Thanks to @mme & @JackDanger for their contributions
- Restructure extension C code
- Add table/filter columns support for CTEs
- Extract views as tables from
CREATE
/REFRESH VIEW
- Refactor code using generic treewalker
- fingerprint: Normalize
IN
lists - param_refs: Fix length attribute in result
- Query fingerprinting
- Filter columns (aka columns referenced in a query's
WHERE
clause) - Parameter references: Returns all
$1
/$2
/etc like references in the query with their location - Remove dependency on active_support
- Fix compilation of C extension
- Fix gemspec
- Speed up build time by only building necessary objects
- PostgreSQL 9.4 parser
See git commit log for previous releases.