Mergeable rustdoc cross-crate info #3662

EtomicBomb · 2024-06-20T20:44:52Z

This RFC is to enable merging cross-crate information (like the search index, index.html, source files index) from separate output directories, so that rustdoc can run in parallel in non-cargo build systems.

Currently, rustdoc updates these cross-crate files with the information from a new crate by reading the old version, adding the information for the current crate, and writing the same file back. This causes issues in build systems that disallow multiple build actions targeting the same output files. Cargo supports cross-crate information because it doesn't have this restriction, Buck2 and Bazel do not support cross-crate information, and Fuchsia has a doc step that runs rustdoc serially on 2700 crates.

Supporting parallel rustdoc with CCI as a native step in Buck2, Bazel, and GN (Fuchsia + Chrome) would require mergable cross-crate information, which is what this proposal is about. It adds new flags, --include-info-json, --write-info-json, --merge=none|read-write|write-only and --include-rendered-docs, to opt-in to these new features.

GuillaumeGomez · 2024-06-20T21:48:58Z

cc @rust-lang/rustdoc

aDotInTheVoid · 2024-06-21T11:46:53Z

Zulip Discussion, WIP Implementation

text/0000-mergeable-rustdoc-cross-crate-info.md

jsha · 2024-06-22T15:44:10Z

I'm very in favor of this! In addition to the parallel building use case, I think it will make it much more feasible to support documentation of namespaced crates as a group. We discussed this a bit on Zulip last month. And merging the parts after doc generation is generally a nicer design.

For implementation details: I'd like for the cross crate info to be written in JSON, not JS, and for the merge step to generate the JS. The JS as it stands is just a very thin callback wrapper around JSON. We do this so docs can be loaded from file:// URLs (where fetching JSON files is not supported). But we don't need the merge step to understand the JS.

Some bikeshedding about the flags: I don't like the names --write-rendered-cci and --read-rendered-cci because they're long, because they contain an acronym that will be opaque to all but rustdoc implementers (cci), and because it's not clear to me that all four combinations of (true, false) values are valid or useful.

It seems to me that the merge step is effectively a subcommand, except that since rustdoc doesn't have subcommands it's triggered by a combination of --write-rendered-cci and --fetch-parts. Is that right?

Could you add a section to the RFC about compatibility? For compatibility on disk, I'm assuming we won't support merging cross crate info that was generated by disparate versions of rustdoc.

As an alternate design, what about this:

unconditionally start generating a crate-info.json in the root of any crate's doc build.
crate-info contains all the different kinds of potentially cross crate info (src files, implementers, etc)
merge step reads crate-info.json from N different crates and generates .js in the appropriate places
New flag: --merge=[auto|none|<list of paths>]
auto is the default value, and causes the equivalent of the current behavior. At the end of each run, rustdoc takes a lock and reads/writes from the out dir to do merge.
none is the future default. Individual rustdoc invocations do no merging; it is up to the build system to do a final merge step.
- <list of paths> causes the merge step to be invoked and no other work to be done. Causes most other rustdoc flags to be ignored.

notriddle · 2024-06-22T16:54:02Z

I would prefer not to put large files in the target/doc folder that aren't actually used by the frontend. People should be able to upload that folder to their webhost without having to filter out subfolders to avoid wasting space.

Instead, I'd design the CLI with two parameters:

--write-info-json=PATH: When supplied, the crate info json (search index, search desc shards, search type param names, foreign trait impls, type aliases, and source code file tree) gets written to a JSON file supplied by PATH. Cargo will use a path that's somewhere in target/, but not in the target/doc folder, so publishers don't accidentally upload files to their web host that don't actually need to be there. Other build systems can use a location that happens to be convenient.
--include-info-json=PATH1,PATH2,...: When --write-info-json isn't supplied, the info files are written the same way they are now. If --include-info-json is supplied, rustdoc will additionally include the data from the supplied paths.

EtomicBomb · 2024-06-22T19:28:49Z

@jsha, I love that there are more use cases than we initially anticipated!

I agree that the intermediate cross-crate info should be written in JSON, and for the merge step to generate JS (and HTML where applicable)! The current proposal / WIP writes intermediate files as JSON (src-files-js, search-index-js, search-desc, all-crates, crates-index, type-impl, trait-impl) with --parts-out-dir=PATH. It renders the JS + HTML versions upon --write-rendered-cci (default). It reads to and appends to CCI in the doc root upon --read-rendered-cci (default), to match the current behavior of rustdoc. I'll put the JSON files into a single crates-info.json and change the names of the options so they make more sense.

Here are what the combinations of --read-rendered-cci and --write-rendered-cci meant. Yes, only some of them are useful, so I agree that a --merge=[] will make more sense.

--read-rendered-cci=false --write-rendered-cci=false --parts-out-dir=target/doc.parts: write the pre-rendered cross-crate information parts files. Merge these with another crates version with another invocation of rustdoc --fetch-parts <...>
--read-rendered-cci=false --write-rendered-cci=false (leaving out --parts-out-dir): Does not read or write any cross-crate information. The user is only interested in item docs.
--read-rendered-cci=false --write-rendered-cci=true: write the cross-crate information for the current crate to src-files.js, search-index.js, type.impl/**/*.js, ignoring the files that are already there. This user may have a hermetic build system that monitors reads.
--read-rendered-cci=true --write-rendered-cci=false: not useful - we only read CCI to append to existing CCI
--read-rendered-cci=true --write-rendered-cci=true: rustdoc's current behavior

I initially conceptualized merging as a subcommand, but you could also see it as an incremental enhancement: if you're calling rustdoc on one of your binaries or libraries, you still might want to link-in the crate-info.json for other crates, without an additional rustdoc invocation. For pure merges though, we could make something that looks more like a merge subcommand: maybe a third party tool that effectively merges into a dummy crate with an empty lib.rs?

It seems like --include-info-json=PATH1,PATH2,... is similar to what this proposal means by --fetch-parts. And --write-info-json=PATH is like --parts-out-dir, and that crate-info.json is like the concatenation of {src-files-js, search-index-js, search-desc, all-crates, crates-index, type-impl, trait-impl}.

Based off of @notriddle and @jsha's feedback, I will update the RFC + WIP with these items:

Replace {src-files-js, search-index-js, search-desc, all-crates, crates-index, type-impl, trait-impl} into a single file crate-info.json
Rename --fetch-parts to the suggested --include-info-json
Rename --parts-out-dir to the suggested --write-info-json
Create a flag --merge=auto to replace --read-rendered-cci=true --write-rendered-cci=true
Create a flag --merge=none to replace --read-rendered-cci=false --write-rendered-cci=false
Create a flag --merge=no-read to replace --read-rendered-cci=false --write-rendered-cci=true
Describe the compatibility promises: crate-info.json is only stable within a version of rustdoc. The CCI after this proposal is not byte-for-byte identical with those produced by previous versions of rustdoc,but should have the same effect when executed by a browser (modulo sorting order).

I'll have to work on these suggestions on Monday, but definitely give me more feedback. Thank you for engaging with this proposal!

EtomicBomb · 2024-06-25T00:04:29Z

Updated the RFC with the items mentioned in my previous comment. I chose the name --merge=write-only instead of --merge=no-read

bjorn3 · 2024-06-25T13:13:36Z

text/0000-mergeable-rustdoc-cross-crate-info.md

+rustdoc -Z unstable-options --crate-name=i --crate-type=lib --edition=2021 --enable-index-page --out-dir=i/target/doc --extern-html-root-url s=$(MERGED) --extern-html-root-url t=$(MERGED) --extern-html-root-url i=$(MERGED) --merge=write-only --include-info-json t=t/target/doc.parts --include-info-json s=s/target/doc.parts --extern t=t/target/libt.rmeta --extern s=s/target/libs.rmeta -L t/target i/src/lib.rs
+```
+
+Merge the docs with `cp`. This can be avoided if `--out-dir=$(MERGED)` is used for all of the rustdoc calls. We copy here to illustrate that documenting `s` is independent of documenting `t`, and could happen on separate machines.


Would it make sense for the rustdoc merge invocation to do all copying too? This would allow the per-crate rustdoc invocations to avoid copying all shared resources in their own doc dirs, saving disk space. And in general it seems like it would make things easier for the build system.

cp is the stand-in for the step that takes docs from external crates and collects them in a single web root. Depending on the location of the external crates, this step may look very different:

It may do nothing; Cargo generates the docs in a single doc/ folder.

cp or mv files from one location on the filesystem to another. This is shown in the example in the RFC.

scp, rsync, wget --recursive, a proprietary protocol, etc. This RFC is intended to enable CCI in crates that have been documented on separate machines.

Because of the wide variety of locations that external docs may be, I believe that the build system is best positioned to collect them into a single location.

When you pass the --merge=none flag to document the external crates, the shared resources are not generated. Only crate-info.json (--write-info-json) and the item docs will be generated for these crates. The shared resources are only generated when you pass --merge=auto (default) or --merge=write-only, which is only provided when documenting the index or root crate. There are not any redundant files to bloat the disk space.

Then please update so that it is mentioned that rustdoc will handle the copying internally as we don't want to rely on external tools.

Okay, I will update it so that rustdoc does the copying.

…-info.json

EtomicBomb · 2024-06-28T00:25:37Z

I updated the RFC + WIP implementation so that --include-info-json and --write-info-json take a path directly to a crate-info.json file.

The existing tests/rustdoc* suite passes my WIP implementation. That hopefully means that no changes will need to be made to tools that use rustdoc.

I hacked compiletest to pass --merge=none + --write-info-json to the auxiliary crates and --merge=write-only + --include-info-json to the main crate for every test. The tests/rustdoc* (except rustdoc-gui - have not run these) tests pass under this mode too. These tests were for my confidence, as I'm not looking to make any changes to compiletest.

djkoloski · 2024-07-15T20:42:21Z

If there are no further blockers for this, can we put this RFC into FCP?

GuillaumeGomez · 2024-07-16T08:32:44Z

text/0000-mergeable-rustdoc-cross-crate-info.md

+
+# Guide-level explanation
+
+In this example, there is a crate `t` which defines a trait `T`, and a crate `s` which defines a struct `S` that implements `T`. Our goal in this demo is for `S` to appear as in implementer in `T`'s docs, even if `s` and `t` are documented independently. This guide will be assuming that we want a crate `i` that serves as our documentation index. See the Unresolved questions section for ideas that do not require an index crate.


Suggested change

In this example, there is a crate `t` which defines a trait `T`, and a crate `s` which defines a struct `S` that implements `T`. Our goal in this demo is for `S` to appear as in implementer in `T`'s docs, even if `s` and `t` are documented independently. This guide will be assuming that we want a crate `i` that serves as our documentation index. See the Unresolved questions section for ideas that do not require an index crate.

In this example, there is a crate `t` which defines a trait `T`, and a crate `s` which defines a struct `S` that implements `T`. Our goal in this demo is for `S` to appear as an implementer in `T`'s docs, even if `s` and `t` are documented independently. This guide will be assuming that we want a crate `i` that serves as our documentation index. See the Unresolved questions section for ideas that do not require an index crate.

GuillaumeGomez · 2024-07-16T08:53:35Z

text/0000-mergeable-rustdoc-cross-crate-info.md

+        │           └── lib.rs.html
+        ├── doc.parts
+        │   └── t
+│       │       └── crate-info.json


Sorry it bugs me. 😆

Suggested change

│ │ └── crate-info.json

│ └── crate-info.json

GuillaumeGomez · 2024-07-16T08:57:53Z

text/0000-mergeable-rustdoc-cross-crate-info.md

+
+When `read_rendered_cci` is active, rustdoc will look in the `--out-dir` for rendered cross-crate info files. These files will be used as the base. Any new parts that rustdoc generates with its current invocation and any parts fetched with `include-info-json` will be appended to these base files. When it is disabled, the cross-crate info files start empty and are populated with the current crate's info and any crates fetched with `--include-info-json`.
+
+* `--merge=auto` (`read_rendered_cci && write_rendered_cci`) is the default, and reflects the current behavior of rustdoc. 


I'm not a big fan of auto. Why not read-write instead? Would be more explicit (took me a while to get why it was different to write-only 😅).

That sounds good!

GuillaumeGomez · 2024-07-16T08:59:39Z

text/0000-mergeable-rustdoc-cross-crate-info.md

+
+This RFC does not alter previous compatibility guarantees made about the output of rustdoc. In particular it does not stabilize the presence of the rendered cross-crate information files, their content, or the HTML generated by rustdoc.
+
+In the same way that the [rustdoc HTML output is unstable](https://rust-lang.github.io/rfcs/2963-rustdoc-json.html#:~:text=The%20HTML%20output%20of%20rustdoc,into%20a%20different%20format%20impractical), the contents of the `crate-info.json` will be considered unstable. Between versions of rustdoc, breaking changes to the content of `crate-info.json` should be expected. Only the presence of a `crate-info.json` file is promised, under `--write-info-json`. Merging cross-crate information generated by disparate versions of rustdoc is not supported.


Suggested change

In the same way that the [rustdoc HTML output is unstable](https://rust-lang.github.io/rfcs/2963-rustdoc-json.html#:~:text=The%20HTML%20output%20of%20rustdoc,into%20a%20different%20format%20impractical), the contents of the `crate-info.json` will be considered unstable. Between versions of rustdoc, breaking changes to the content of `crate-info.json` should be expected. Only the presence of a `crate-info.json` file is promised, under `--write-info-json`. Merging cross-crate information generated by disparate versions of rustdoc is not supported.

In the same way that the [rustdoc HTML output is unstable](https://rust-lang.github.io/rfcs/2963-rustdoc-json.html#:~:text=The%20HTML%20output%20of%20rustdoc,into%20a%20different%20format%20impractical), the content of the `crate-info.json` will be considered unstable. Between versions of rustdoc, breaking changes to the content of `crate-info.json` should be expected. Only the presence of a `crate-info.json` file is promised, under `--write-info-json`. Merging cross-crate information generated by disparate versions of rustdoc is not supported.

You say:

Merging cross-crate information generated by disparate versions of rustdoc is not supported.

But in the description of the content of the file, you don't mention that the version of rustdoc is stored or how it's stored.

Do you have a suggestion about how to version the crate-info.json files? We could put the semver as a key in the JSON and accept an --info-json-version flag. Rustdoc would fail if it is passed an --include-info-json with an incompatible file.

That's a tricky question. Currently the rustdoc JSON output has a simple counter but it's something we're thinking about changing, but we didn't find a new versioning system yet (you can read about it here).

So really not sure what to pick, but this is definitely a blocker for this RFC.

I think that an incremental single-digit version number makes the most sense. The RFC only discusses the case where crate-info.json files are written by rustdoc, and consumed by the same version rustdoc, so it doesn't seem important to detail what constitutes a breaking change. Also the content of the files is tied to the internal details of rustdoc (including the details of the search index). I'll make rustdoc write an incremental single-digit version to the crate-info.json file, and rustdoc will fail if it's called with an --include-info-json pointing to a crate-info.json with a version that does not match its internal version.

text/0000-mergeable-rustdoc-cross-crate-info.md

* remove 'Unresolved: index crate?' from the unresolved questions section * remove the requirement on the index crate from various sections

EtomicBomb · 2024-08-16T02:28:15Z

I updated the RFC to resolve in favor of not needing an index crate to merge docs in a workspace, as explained above.

* Adds `--merge=shared|none|finalize` flags * Adds `--parts-out-dir=<crate specific directory>` for `--merge=none` to write cross-crate info file for a single crate * Adds `--include-parts-dir=<previously specified directory>` for `--merge=finalize` to write cross-crate info files * Longer crate names in previous cross-crate-info tests * Unit tests for above

jsha · 2024-08-20T14:28:20Z

We should expect the size of crate-info to grow in proportion with the size of the crate's docs, since it includes all cross-crate trait implementations and the entire search index. For doc indexes with many crates, we should expect this file to be very large. @dtolnay describes how a search index for all of the crates in the doc index maintained by his team would be several gigabytes in size.

That's the merged search index, though - not each crate's search index.

I think that build systems would strongly prefer for intermediate files to be separate from the output files.

This is a good point. It would perhaps make sense to consider all of the regular rustdoc outputs to be intermediates, which are merged in the merge step. But maybe that introduces an excessive number of filesystem operations? At any rate, in the interests of unblocking the work I'm willing to concede this point and say the extra flag is okay.

jsha

Added my approval on the FCP comment. Also one editorial nit below.

jsha · 2024-08-20T14:28:57Z

text/0000-mergeable-rustdoc-cross-crate-info.md

@@ -33,26 +35,20 @@ More details are in the Reference-level explanation.
 * `--include-parts-dir=path/to/doc.parts/<crate-name>`: Include cross-crate information from this previously written `doc.parts` directories into a collection that will be written by the current invocation of rustdoc. May only be provided with `--merge=finalize`. May be provided any number of times.
 * `--merge=none`: Do not write cross-crate information to the `--out-dir`. The flag `--parts-out-dir` may instead be provided with the destination of the current crate's cross-crate information parts.
 * `--merge=shared` (default): Append information from the current crate to any info files found in the `--out-dir`.
-* `--merge=finalize`: Write cross-crate information from the current crate and any crates included via `--include-parts-dir` to the `--out-dir`, overwriting conflicting files.
+* `--merge=finalize`: Write cross-crate information from the current crate and any crates included via `--include-parts-dir` to the `--out-dir`, overwriting conflicting files. This flag may be used with or without an input crate root, as it can links existing docs.


Suggested change

* `--merge=finalize`: Write cross-crate information from the current crate and any crates included via `--include-parts-dir` to the `--out-dir`, overwriting conflicting files. This flag may be used with or without an input crate root, as it can links existing docs.

* `--merge=finalize`: Write cross-crate information from the current crate and any crates included via `--include-parts-dir` to the `--out-dir`, overwriting conflicting files. This flag may be used with or without an input crate root, as it can link existing docs.

typo in --merge=finalize description: can links

modularize rustdoc's write_shared Refactor src/librustdoc/html/render/write_shared.rs to reduce code duplication, adding unit tests * Extract + unit test code for sorting and rendering JSON, which is duplicated 9 times in the current impl * Extract + unit test code for encoding JSON as single quoted strings, which is duplicated twice in the current impl * Unit tests for cross-crate information file formats * Generic interface to add new kinds of cross-crate information files in the future * Intended to match current behavior exactly, except for a merge info comment it adds to the bottom of cci files * This PR is intended to reduce the review burden from my [mergeable rustdoc rfc](rust-lang/rfcs#3662) implementation PR, which is a [small commit based on this branch](https://github.com/EtomicBomb/rust/tree/rfc). This code is agnostic to the RFC and does not include any of the flags discussed there, but cleanly enables the addition of these flags in the future because it is more modular

aDotInTheVoid

Checked my FCP box, but have a few editorial comments and requests for clarification (where the answer doesn't make a difference to weather I'm happy with this, as long as it get's written down in the RFC itself)

text/0000-mergeable-rustdoc-cross-crate-info.md

aDotInTheVoid · 2024-08-20T23:36:33Z

text/0000-mergeable-rustdoc-cross-crate-info.md

+
+In the same way that the [rustdoc HTML output is unstable](https://rust-lang.github.io/rfcs/2963-rustdoc-json.html#:~:text=The%20HTML%20output%20of%20rustdoc,into%20a%20different%20format%20impractical), the content of `doc.parts` will be considered unstable. Between versions of rustdoc, breaking changes to the content of `doc.parts` should be expected. Only the presence of a `doc.parts` directory is promised, under `--parts-out-dir`. Merging cross-crate information generated by disparate versions of rustdoc is not supported. To detect whether `doc.parts` is compatible, rustdoc includes a version number in these files (see New directory: `doc.parts`).
+
+The implementation of the RFC itself is designed to produce only minimal changes to cross-crate info files and the HTML output of rustdoc. Exhaustively, the implementation is allowed to 


Editorial: What is this talking about? There was a linked draft implementation in the PR, but that was on the master branch of your fork, which now doesn't have it. It's worth clairifying that this is your WIP implementation (unless it isn't), and linking to it.

I removed the reference. It was to a different RFC (#2963), but I don't think the reference is relevant anymore.

text/0000-mergeable-rustdoc-cross-crate-info.md

* full sentences in intro * Clarify Cargo's expected use of `--merge=none|shared|finalize` * Explain why you would use multiple rustdoc invocations with the same --out-dir and merge=none * Remove justification for why rustdoc output can be unstable in general * Permalink specific fuchsia commit

camelid

@rfcbot reviewed
@rfcbot resolve complexity

Thanks! I really like the way this has turned out. I just have one non-essential suggestion.

camelid · 2024-08-21T18:36:07Z

text/0000-mergeable-rustdoc-cross-crate-info.md

+* `--merge=none`: Do not write cross-crate information to the `--out-dir`. The flag `--parts-out-dir` may instead be provided with the destination of the current crate's cross-crate information parts.
+* `--parts-out-dir=path/to/doc.parts/<crate-name>`: Write cross-crate linking information to the given directory (only usable with the `--merge=none` mode). This information allows linking the current crate's documentation with other documentation at a later rustdoc invocation.
+* `--include-parts-dir=path/to/doc.parts/<crate-name>`: Include cross-crate information from this previously written `doc.parts` directories into a collection that will be written by the current invocation of rustdoc. May only be provided with `--merge=finalize`. May be provided any number of times.
+* `--merge=shared` (default): Append information from the current crate to any info files found in the `--out-dir`.


Since we plan to deprecate --merge=shared anyway, what do you think about not adding it in the first place? --merge already has to be an optional flag to avoid breaking changes.

Uniformity and ease of documentation. It's the same reason you're allowed to write edition = 2015, even though that's the default if you don't specify anything.

Fair enough. I guess we won't ever be able to remove the shared implementation anyway, so it doesn't matter. And if we somehow did find a way to remove it while maintaining the outward behavior, that wouldn't be a breaking change anyway since it's an implementation detail.

Why do you say we won't ever be able to remove the shared implementation? I've been assuming that after updating Cargo and docs.rs, and providing a decently long deprecation period we could remove it.

Hmm, I'm trying to remember what I was thinking ^^. I think I was worried that there might be existing build systems or other tools that were depending on the shared representation, but we never guaranteed stability of it. So perhaps it would be fine to remove after all.

text/0000-mergeable-rustdoc-cross-crate-info.md

camelid · 2024-08-21T18:47:21Z

@rfcbot reviewed
@rfcbot resolve complexity

rfcbot · 2024-08-21T18:47:26Z

🔔 This is now entering its final comment period, as per the review above. 🔔

Change sentence reading "The WIP may change the sorting order..." to "The implementation may change the sorting order..." Co-authored-by: Noah Lev <camelidcamel@gmail.com>

modularize rustdoc's write_shared Refactor src/librustdoc/html/render/write_shared.rs to reduce code duplication, adding unit tests * Extract + unit test code for sorting and rendering JSON, which is duplicated 9 times in the current impl * Extract + unit test code for encoding JSON as single quoted strings, which is duplicated twice in the current impl * Unit tests for cross-crate information file formats * Generic interface to add new kinds of cross-crate information files in the future * Intended to match current behavior exactly, except for a merge info comment it adds to the bottom of cci files * This PR is intended to reduce the review burden from my [mergeable rustdoc rfc](rust-lang/rfcs#3662) implementation PR, which is a [small commit based on this branch](https://github.com/EtomicBomb/rust/tree/rfc). This code is agnostic to the RFC and does not include any of the flags discussed there, but cleanly enables the addition of these flags in the future because it is more modular

rfcbot · 2024-08-31T18:48:52Z

The final comment period, with a disposition to merge, as per the review above, is now complete.

As the automated representative of the governance process, I would like to thank the author for their work and everyone else who contributed.

This will be merged soon.

camelid · 2024-09-21T21:16:54Z

We should discuss this before stabilization. In the meantime, this RFC needs to get actually merged, so I'll do that now. :)

camelid · 2024-09-21T21:55:07Z

Huzzah! The @rust-lang/rustdoc team has decided to accept this RFC.

To track further discussion, subscribe to the tracking issue here:
rust-lang/rust#130676

EtomicBomb added 8 commits June 19, 2024 02:52

add mergable_rustdoc_cross_crate_info

542dcbb

generates no files by default

fdc1b70

fix typos

8c78a93

merged -> rendered in flags

08ac8a2

fix typos

ef10ed5

whitespace and clear lines

4994251

typo

0f01c9b

remove work in progress

7db0a4a

ehuss added the T-rustdoc Relevant to rustdoc team, which will review and decide on the RFC. label Jun 20, 2024

EtomicBomb mentioned this pull request Jun 21, 2024

Rules to run rustdoc on multiple crates bazelbuild/rules_rust#1837

Open

P1n3appl3 suggested changes Jun 21, 2024

View reviewed changes

address comments from P1n3appl3

23e34e0

EtomicBomb added 3 commits June 24, 2024 18:15

changes in response to jsha and notriddle

97cf6c6

typos

3326b10

nits

cbd62bb

bjorn3 reviewed Jun 25, 2024

View reviewed changes

--include-info-json and --write-info-json take path directly to crate…

eba7fc4

…-info.json

GuillaumeGomez reviewed Jul 16, 2024

View reviewed changes

text/0000-mergeable-rustdoc-cross-crate-info.md Outdated Show resolved Hide resolved

resolve in favor of no index crate

ac9208c

* remove 'Unresolved: index crate?' from the unresolved questions section * remove the requirement on the index crate from various sections

EtomicBomb added a commit to EtomicBomb/rust that referenced this pull request Aug 16, 2024

changes from rfc rust-lang/rfcs#3662

91f4a48

jsha approved these changes Aug 20, 2024

View reviewed changes

fix typo identified by jsha

e66bb63

typo in --merge=finalize description: can links

aDotInTheVoid reviewed Aug 20, 2024

View reviewed changes

text/0000-mergeable-rustdoc-cross-crate-info.md Outdated Show resolved Hide resolved

camelid approved these changes Aug 21, 2024

View reviewed changes

rfcbot added the final-comment-period Will be merged/postponed/closed in ~10 calendar days unless new substational objections are raised. label Aug 21, 2024

rfcbot removed the proposed-final-comment-period Currently awaiting signoff of all team members in order to enter the final comment period. label Aug 21, 2024

replace WIP -> implementation, per camelid

e1b5e5b

Change sentence reading "The WIP may change the sorting order..." to "The implementation may change the sorting order..." Co-authored-by: Noah Lev <camelidcamel@gmail.com>

rfcbot added finished-final-comment-period The final comment period is finished for this RFC. and removed final-comment-period Will be merged/postponed/closed in ~10 calendar days unless new substational objections are raised. labels Aug 31, 2024

rfcbot added the to-announce label Aug 31, 2024

This was referenced Sep 21, 2024

Tracking Issue for mergeable rustdoc cross-crate info rust-lang/rust#130676

Open

RFC 3662: Mergeable rustdoc cross-crate info #3699

Merged

camelid merged commit e1b5e5b into rust-lang:master Sep 21, 2024

This was referenced Nov 1, 2024

rustdoc: prefer showing enum variants as written rust-lang/rust#131975

Open

rustdoc: Consider adding CCI for the HIR expressions of public constants rust-lang/rust#132443

Open


		# Guide-level explanation

		In this example, there is a crate `t` which defines a trait `T`, and a crate `s` which defines a struct `S` that implements `T`. Our goal in this demo is for `S` to appear as in implementer in `T`'s docs, even if `s` and `t` are documented independently. This guide will be assuming that we want a crate `i` that serves as our documentation index. See the Unresolved questions section for ideas that do not require an index crate.

	│ │ └── crate-info.json
	│ └── crate-info.json


		When `read_rendered_cci` is active, rustdoc will look in the `--out-dir` for rendered cross-crate info files. These files will be used as the base. Any new parts that rustdoc generates with its current invocation and any parts fetched with `include-info-json` will be appended to these base files. When it is disabled, the cross-crate info files start empty and are populated with the current crate's info and any crates fetched with `--include-info-json`.

		* `--merge=auto` (`read_rendered_cci && write_rendered_cci`) is the default, and reflects the current behavior of rustdoc.


		This RFC does not alter previous compatibility guarantees made about the output of rustdoc. In particular it does not stabilize the presence of the rendered cross-crate information files, their content, or the HTML generated by rustdoc.

		In the same way that the [rustdoc HTML output is unstable](https://rust-lang.github.io/rfcs/2963-rustdoc-json.html#:~:text=The%20HTML%20output%20of%20rustdoc,into%20a%20different%20format%20impractical), the contents of the `crate-info.json` will be considered unstable. Between versions of rustdoc, breaking changes to the content of `crate-info.json` should be expected. Only the presence of a `crate-info.json` file is promised, under `--write-info-json`. Merging cross-crate information generated by disparate versions of rustdoc is not supported.

	* `--merge=finalize`: Write cross-crate information from the current crate and any crates included via `--include-parts-dir` to the `--out-dir`, overwriting conflicting files. This flag may be used with or without an input crate root, as it can links existing docs.
	* `--merge=finalize`: Write cross-crate information from the current crate and any crates included via `--include-parts-dir` to the `--out-dir`, overwriting conflicting files. This flag may be used with or without an input crate root, as it can link existing docs.


		In the same way that the [rustdoc HTML output is unstable](https://rust-lang.github.io/rfcs/2963-rustdoc-json.html#:~:text=The%20HTML%20output%20of%20rustdoc,into%20a%20different%20format%20impractical), the content of `doc.parts` will be considered unstable. Between versions of rustdoc, breaking changes to the content of `doc.parts` should be expected. Only the presence of a `doc.parts` directory is promised, under `--parts-out-dir`. Merging cross-crate information generated by disparate versions of rustdoc is not supported. To detect whether `doc.parts` is compatible, rustdoc includes a version number in these files (see New directory: `doc.parts`).

		The implementation of the RFC itself is designed to produce only minimal changes to cross-crate info files and the HTML output of rustdoc. Exhaustively, the implementation is allowed to

Mergeable rustdoc cross-crate info #3662

Mergeable rustdoc cross-crate info #3662

Conversation

EtomicBomb commented Jun 20, 2024 • edited by camelid Loading

GuillaumeGomez commented Jun 20, 2024

aDotInTheVoid commented Jun 21, 2024

jsha commented Jun 22, 2024

notriddle commented Jun 22, 2024 • edited Loading

EtomicBomb commented Jun 22, 2024 • edited Loading

EtomicBomb commented Jun 25, 2024 • edited Loading

Choose a reason for hiding this comment

EtomicBomb Jun 25, 2024 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

EtomicBomb commented Jun 28, 2024 • edited Loading

djkoloski commented Jul 15, 2024

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

Choose a reason for hiding this comment

EtomicBomb Jul 16, 2024 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

EtomicBomb commented Aug 16, 2024

jsha commented Aug 20, 2024

jsha left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

aDotInTheVoid left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

EtomicBomb Aug 21, 2024 • edited Loading

Choose a reason for hiding this comment

camelid left a 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

Choose a reason for hiding this comment

Choose a reason for hiding this comment

camelid commented Aug 21, 2024

rfcbot commented Aug 21, 2024

rfcbot commented Aug 31, 2024

camelid commented Sep 21, 2024

camelid commented Sep 21, 2024

EtomicBomb commented Jun 20, 2024 •

edited by camelid

Loading

notriddle commented Jun 22, 2024 •

edited

Loading

EtomicBomb commented Jun 22, 2024 •

edited

Loading

EtomicBomb commented Jun 25, 2024 •

edited

Loading

EtomicBomb Jun 25, 2024 •

edited

Loading

EtomicBomb commented Jun 28, 2024 •

edited

Loading

EtomicBomb Jul 16, 2024 •

edited

Loading

EtomicBomb Aug 21, 2024 •

edited

Loading