Add doc aliases for python and Java collections #81989

jyn514 · 2021-02-11T04:37:32Z

No description provided.

rust-highfive · 2021-02-11T04:37:35Z

r? @Mark-Simulacrum

(rust-highfive has picked a reviewer for you, use r? to override)

library/std/src/collections/hash/map.rs

library/std/src/collections/hash/set.rs

camelid · 2021-02-14T21:42:03Z

library/alloc/src/collections/btree/map.rs

@@ -137,6 +137,7 @@ pub(super) const MIN_LEN: usize = node::MIN_LEN_AFTER_SPLIT;
 /// *stat += random_stat_buff();
 /// ```
 #[stable(feature = "rust1", since = "1.0.0")]
+#[doc(alias = "map")]


Given that you're adding map to BTreeMap, do you want to add set to BTreeSet? Either way they should show up in the results though, so not really necessary.

Good idea, thanks. Added.

Either way they should show up in the results though, so not really necessary.

This will make it show up in the top, right now it's a long way down.

camelid · 2021-02-14T21:42:27Z

library/alloc/src/vec/mod.rs

@@ -1486,6 +1488,7 @@ impl<T, A: Allocator> Vec<T, A> {
    /// ```
    #[inline]
    #[stable(feature = "rust1", since = "1.0.0")]
+    #[doc(alias = "append")]


I think this makes sense to add, but it is unfortunate that it will pollute results for Vec::append.

Do you know if the alias or Vec::append will be shown first in the results?

Also could potentially be worth adding a line to Vec::append's docs like:

Not to be confused with Vec::push. Vec::append is different from .append in Python.

Do you know if the alias or Vec::append will be shown first in the results?

Aliases should always be shown last in the search results, I'd say, if they are not already.

Vec::append is different from .append in Python.

I like the idea, but I think adding references to other languages in the documentation for the Rust function is going a bit too far. Instead, I'd go for #82000, allowing a comment field for each alias too, where this comment can be written. Then, the search can show it however is best, i.e. something like:

#[doc(alias = "append", lang = "Python", comment = "Not to be confused with `Vec::push`. `Vec::append` is different from `.append` in Python." )]

What do you think? I can add it to the RFC.

Sure, seems reasonable. I'll drop append from this PR.

Aliases should always be shown last in the search results, I'd say, if they are not already.

That seems really strange, why do you say that? That puts them behind very unrelated things because of fuzzy search.

That seems really strange, why do you say that? That puts them behind very unrelated things because of fuzzy search.

Ah, yeah, there are too many results [*]. What about before the fuzzy ones? i.e.:

Exact matches (non-aliases)

Exact matches (aliases)

Fuzzy search (any, assuming it sorts by some "similarity" index)

If we get #82000, then we could even do:

Exact matches (non-aliases)

Exact matches (Rust aliases)

Exact matches (non-Rust aliases)

Fuzzy search (any, assuming it sorts by some "similarity" index)

[*] Why is matching all that for free? Is it due to substrings (e.g. fREm -> fREe) or Hamming distance or something like that? I think for words < 4 in size, it should not match if at least 3 characters match. But anyway, this is off-topic... :-) The general point is that the search results seem like they could be improved (i.e. instead of working around it with the aliases).

Ah, yeah, there are too many results [*]. What about before the fuzzy ones? i.e.:

I like this. Can you open an issue for it? I think it makes a lot of sense.

[*] Why is matching all that for free? Is it due to substrings (e.g. fREm -> fREe) or Hamming distance or something like that? I think for words < 4 in size, it should not match if at least 3 characters match. But anyway, this is off-topic...

Can you open an issue for that too? 😆

Done! #82131 and #82132.

Mark-Simulacrum · 2021-02-20T21:17:18Z

Hm, it seems not bad necessarily but I'm a bit worried about eventually everything in std gaining 3-4 different aliases, some conflicting, depending on perspective. It seems like we should at least put a comment next to each one for origin language -- I wonder if aliasing as "python dict" would be a bad idea.

cc @rust-lang/libs, curious on thoughts.

jyn514 · 2021-02-20T21:53:48Z

It seems like we should at least put a comment next to each one for origin language -- I wonder if aliasing as "python dict" would be a bad idea.

See also #82000.

m-ou-se · 2021-02-20T22:11:10Z

Hm, it seems not bad necessarily but I'm a bit worried about eventually everything in std gaining 3-4 different aliases, some conflicting, depending on perspective. It seems like we should at least put a comment next to each one for origin language -- I wonder if aliasing as "python dict" would be a bad idea.

Yeah I feel like adding all the equivalents of everything in popular other langues quickly gets out of hand. I think I'd prefer to only add those where the word is common enough that it's often used as a regular noun/verb, like "memcpy it to some newly malloc'd buffer" or "put the pairs in a map". So for this specific PR, I'd say map, dict, and set are good examples, but I don't think we should add ArrayList.

I also think we should be extra careful with names for things that already have another related meaning in Rust. For example, depending on the languages someone might already be familiar with, the difference between an array and Vec could be pretty confusing. Adding an alias like Array to Vec might make that even more confusing. In those cases, I rather add some notes to the array and Vec documentation linking each other.

BurntSushi · 2021-02-21T15:52:10Z

I agree with @m-ou-se and would strongly opposed adding specific aliases like ArrayList to the docs. In addition to such a thing getting quickly out of hand, it could also be misleading. If we start aliasing specific data structures from other ecosystems, while there are assuredly similarities (perhaps even almost identical), there may also be differences between them. Those differences, depending on one's perspective, may or may not make aliases surprising to folks.

Using abstractions like "map" and "set" as aliases seems fine to me. Nobody owns those words.

I'd say "dict" is somewhat border line since it's very strongly associated with one ecosystem in particular. Ecosystems use lots of different names for their map types. "associative array" (PHP) and "hash" (Ruby) come to mind. If we add "dict," should we also add those? On the other hand, at least "associative array" and "dict" are fairly abstract words on their own. So I'm torn on those in particular.

jyn514 · 2021-02-23T05:32:58Z

I removed ArrayList and append, changed Map to map (not that it matters, but it means it looks like the concept rather than the interface from Java), and added associated array wherever map is used. I added map/dict/associative array to BTreeMap. I didn't add hash because the first search result is currently for https://doc.rust-lang.org/stable/std/hash/index.html, which I think is a better fit. That module doesn't currently link to HashMap; I'll make a follow-up PR adding that.

ojeda · 2021-02-23T07:25:25Z

dictionary would be a bit better (if typing dict makes the entry appear too).

KamilaBorowska · 2021-02-26T13:17:57Z

I think ArrayList alias is fine myself. It's a very specific name that in all programming languages it is used in it means a contiguous growable array type. Additionally, doc aliases aren't fuzzy searched for, if an user searches for ArrayLis, ArrayLit, ArryList or whatever the documentation simply won't find it, so there is no risk of user finding a Vec when searching for a term that happens to be very similar to ArrayList, it needs to be ArrayList specifically.

If we start aliasing specific data structures from other ecosystems, while there are assuredly similarities (perhaps even almost identical), there may also be differences between them. Those differences, depending on one's perspective, may or may not make aliases surprising to folks.

Well... yes, there are differences between Vec and ArrayList, I just don't think they are particularly meaningful and likely to cause confusion. For instance, resizing strategy is different. Rust's growth factor is 2, Java's growth factor is 1.5. It's a good arguement for considering aliases carefully, especially with data structures more complicated than Vec, it just doesn't really apply here.

I think I'd prefer to only add those where the word is common enough that it's often used as a regular noun/verb, like "memcpy it to some newly malloc'd buffer" or "put the pairs in a map". So for this specific PR, I'd say map, dict, and set are good examples, but I don't think we should add ArrayList.

I'm not seeing myself how ArrayList doesn't meet this criteria. C programmers (and to some extent Rust programmers due to C FFI support) will often use memcpy and malloc as regular words because those identifiers are very often encountered in C programs. Meanwhile Java programmers will often use ArrayList as a regular word because it's a very common identifier in Java programs, similarly to how Rust programmers often use "vector" as a regular word - the Rust documentation even explicitly points that out saying "pronounced 'vector'" in the first line of Vec's documentation.

Anyway, I believe adding ArrayList would help Java programmers who are learning the language. They could try searching for ArrayList if they happen to forget what the dynamically allocated type was called in Rust. Currently if an user searches for ArrayList, this is the resulting page.

Clicking "Try on DuckDuckGo" is not very useful, it will point the user to an array (nope, that's not what the user wanted) and LinkedList (which is honestly worse than pointing nowhere because it kinda works as a vector, but I don't think we want to point beginners to it).

`append` sounds strange because Rust already has append and it does something different. The reason I added it is one of my friends knew from python that it added a single element to the end of an array. When they started learning Rust, rather than looking for pop(), they wrote something like ```rust my_vec.append(&mut vec![x]); ``` which is clearly not ideal. Hopefully this would have pointed them in the right direction.

Add more links between hash and btree collections - Link from `core::hash` to `HashMap` and `HashSet` - Link from HashMap and HashSet to the module-level documentation on when to use the collection - Link from several collections to Wikipedia articles on the general concept See also rust-lang#81989 (comment).

m-ou-se · 2021-04-21T15:15:27Z

We discussed the doc(alias) PRs in the library team meeting, and are not sure what direction we want to take here. At first it seemed good and useful to have things like doc(alias = "memcpy") on the Rust equivalents, but now that more PRs are coming in for more doc aliases, it feels like it's getting a bit out of hand and we need a policy for this first.

We didn't reach a clear conclusion about which kind of aliases we should and shouldn't have, but we did feel like adding more aliases like this is not the way to go.

If someone doesn't know about e.g. Vec and types in list, they're probably better served by a query like "rust list" on their favourite search engine, which will point them to https://doc.rust-lang.org/book/ch08-01-vectors.html and some other resources which can explain them in which ways a Vec is and isn't the kind of list they need.

Similarly for dict or anything else that doesn't map 1:1 from another language to Rust. A web search for "dict rust" will result in blog posts and Stack Overflow answers that explain things about HashMaps, and points them to suggestions for other crates that provide something closer to what they might be looking for. Just applying alias = "dict" to HashMap could be misleading, since they have different properties. (E.g. whether the keep the insertion order.)

We didn't make any final decision, but we do want some kind of clear guideline/policy that we can all agree on before accepting more aliases. Feel free to open an issue on https://github.com/rust-lang/libs-team or ping us on Zulip if you (anyone) want to work on this.

(Edit: For anyone interested, some more details/thoughts on aliases here: #81988 (comment))

jyn514 · 2021-04-21T15:20:42Z

That makes sense to me. I'm not planning to put time into making a long-term policy.

Add more links between hash and btree collections - Link from `core::hash` to `HashMap` and `HashSet` - Link from HashMap and HashSet to the module-level documentation on when to use the collection - Link from several collections to Wikipedia articles on the general concept See also rust-lang/rust#81989 (comment).

jyn514 added T-libs-api Relevant to the library API team, which will review and decide on the PR/issue. A-docs Area: documentation for any part of the project, including the compiler, standard library, and tools labels Feb 11, 2021

rust-highfive assigned Mark-Simulacrum Feb 11, 2021

rust-highfive added the S-waiting-on-review Status: Awaiting review from the assignee but also interested parties. label Feb 11, 2021

jyn514 changed the title ~~Add aliases for python collections~~ Add doc aliases for python collections Feb 11, 2021

jyn514 changed the title ~~Add doc aliases for python collections~~ Add doc aliases for python and Java collections Feb 11, 2021

camelid reviewed Feb 11, 2021

View reviewed changes

library/std/src/collections/hash/map.rs Outdated Show resolved Hide resolved

ojeda reviewed Feb 11, 2021

View reviewed changes

library/std/src/collections/hash/set.rs Show resolved Hide resolved

jyn514 force-pushed the doc-alias-collections branch from eb5e2e4 to 62df50d Compare February 11, 2021 16:53

jyn514 mentioned this pull request Feb 11, 2021

Add doc aliases for a lot of the C standard library #81988

Closed

ojeda mentioned this pull request Feb 11, 2021

doc(alias): should it be a mapping rather than search terms? #82000

Open

jyn514 force-pushed the doc-alias-collections branch from 62df50d to cd04c59 Compare February 14, 2021 16:09

ojeda approved these changes Feb 14, 2021

View reviewed changes

camelid reviewed Feb 14, 2021

View reviewed changes

jyn514 force-pushed the doc-alias-collections branch from cd04c59 to 1a79e76 Compare February 14, 2021 22:14

This comment has been minimized.

Sign in to view

jyn514 force-pushed the doc-alias-collections branch 2 times, most recently from dd41345 to 0188363 Compare February 14, 2021 22:24

Mark-Simulacrum added S-waiting-on-team Status: Awaiting decision from the relevant subteam (see the T-<team> label). and removed S-waiting-on-review Status: Awaiting review from the assignee but also interested parties. labels Feb 20, 2021

jyn514 force-pushed the doc-alias-collections branch from 0188363 to 6591010 Compare February 23, 2021 05:32

jyn514 mentioned this pull request Feb 23, 2021

Add more links between hash and btree collections #82434

Merged

camelid mentioned this pull request Feb 26, 2021

Add some doc aliases to Vec #80432

Closed

This comment has been minimized.

Sign in to view

jyn514 force-pushed the doc-alias-collections branch from 6591010 to a8c8025 Compare March 2, 2021 01:21

crlf0710 added S-waiting-on-team Status: Awaiting decision from the relevant subteam (see the T-<team> label). and removed S-waiting-on-team Status: Awaiting decision from the relevant subteam (see the T-<team> label). labels Apr 9, 2021

jyn514 closed this Apr 21, 2021

jyn514 deleted the doc-alias-collections branch April 21, 2021 15:20

m-ou-se mentioned this pull request Apr 21, 2021

Add take as alias for remove in collection methods #79943

Closed

jyn514 mentioned this pull request Apr 23, 2021

Doc alias for bsearch and qsort #84474

Closed

m-ou-se mentioned this pull request Jun 24, 2021

We need a doc(alias) policy rust-lang/libs-team#18

Closed

est31 mentioned this pull request Jun 3, 2023

add #[doc(alias="flatmap")] to Option::and_then #109093

Merged

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Add doc aliases for python and Java collections #81989

Add doc aliases for python and Java collections #81989

jyn514 commented Feb 11, 2021 •

edited

Loading

rust-highfive commented Feb 11, 2021

camelid Feb 14, 2021

jyn514 Feb 14, 2021

camelid Feb 14, 2021

camelid Feb 14, 2021

camelid Feb 14, 2021

ojeda Feb 14, 2021 •

edited

Loading

jyn514 Feb 14, 2021 •

edited

Loading

jyn514 Feb 14, 2021

ojeda Feb 15, 2021 •

edited

Loading

jyn514 Feb 15, 2021

ojeda Feb 15, 2021

This comment has been minimized.

Mark-Simulacrum commented Feb 20, 2021

jyn514 commented Feb 20, 2021

m-ou-se commented Feb 20, 2021

BurntSushi commented Feb 21, 2021

jyn514 commented Feb 23, 2021

ojeda commented Feb 23, 2021

KamilaBorowska commented Feb 26, 2021 •

edited

Loading

This comment has been minimized.

m-ou-se commented Apr 21, 2021 •

edited

Loading

jyn514 commented Apr 21, 2021

Add doc aliases for python and Java collections #81989

Add doc aliases for python and Java collections #81989

Conversation

jyn514 commented Feb 11, 2021 • edited Loading

rust-highfive commented Feb 11, 2021

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

ojeda Feb 14, 2021 • edited Loading

Choose a reason for hiding this comment

jyn514 Feb 14, 2021 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

ojeda Feb 15, 2021 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

This comment has been minimized.

Mark-Simulacrum commented Feb 20, 2021

jyn514 commented Feb 20, 2021

m-ou-se commented Feb 20, 2021

BurntSushi commented Feb 21, 2021

jyn514 commented Feb 23, 2021

ojeda commented Feb 23, 2021

KamilaBorowska commented Feb 26, 2021 • edited Loading

This comment has been minimized.

m-ou-se commented Apr 21, 2021 • edited Loading

jyn514 commented Apr 21, 2021

jyn514 commented Feb 11, 2021 •

edited

Loading

ojeda Feb 14, 2021 •

edited

Loading

jyn514 Feb 14, 2021 •

edited

Loading

ojeda Feb 15, 2021 •

edited

Loading

KamilaBorowska commented Feb 26, 2021 •

edited

Loading

m-ou-se commented Apr 21, 2021 •

edited

Loading