rustdoc: use more precise relative URLs #83237

notriddle · 2021-03-17T17:28:21Z

This is a fairly large diff, and will probably conflict with #82815 since it reduces (but does not eliminate) the use of the old depth variable.

Instead of using a depth counter and adding "../" to get to the top, this commit makes rustdoc actually compare the path of what it's linking from to the path that it's linking to. This makes the resulting HTML shorter.

Here's a comparison of one of the largest (non-source) files in the Rust standard library docs (about 4% improvement before gzipping).

$ wc -c struct.Wrapping.old.html struct.Wrapping.new.html
2387389 struct.Wrapping.old.html
2298538 struct.Wrapping.new.html

Most if it can be efficiently gzipped away.

$ wc -c struct.Wrapping.old.html.gz struct.Wrapping.new.html.gz
70679 struct.Wrapping.old.html.gz
70050 struct.Wrapping.new.html.gz

But it also makes a difference in the final DOM size, reducing it from 91MiB to 82MiB.

rust-highfive · 2021-03-17T17:28:24Z

r? @jyn514

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

src/librustdoc/clean/types.rs

jyn514

But it also makes a difference in the final DOM size, reducing it from 91MiB to 82MiB.

What is the "DOM size"? Is that like the memory the browser uses the render the page?

I guess before I go much further I should ask: what happens if relative_to is empty? Will it just use links that are relative to the output root? If so I don't feel as bad about bugs, it just means the links are a little more verbose.

src/librustdoc/clean/types.rs

src/librustdoc/clean/mod.rs

src/librustdoc/clean/utils.rs

src/librustdoc/html/format.rs

notriddle · 2021-04-05T19:34:36Z

@jyn514

What is the "DOM size"? Is that like the memory the browser uses the render the page?

Yes, that's exactly what it is. In Firefox, go into the Dev Tools, then pick the Memory tab, then click the Take Snapshot button (camera icon).

what happens if relative_to is empty? Will it just use links that are relative to the output root?

Yes, that's exactly what it does. I've added a test case for this in src/librustdoc/html/tests.rs.

notriddle · 2021-04-05T20:27:13Z

@jyn514

Can you add a test case for that to make sure it doesn't regress?

Done in 53df7fc507bb6e48e1aa6588300f29c0ee4d0437

s.trim_end_matches

Done in 25bc0c0d1364a0b0cd7b8ba4742d91e3f7158b15

jyn514

Fixed in d85b42a

Could you add a test for linking to decl macros?

In some places you use cx.current; in others you choose a custom relative_to. I think the only difference is that some places are in clean, so cx.current isn't set yet. Rather than threading relative_to through everywhere, do you think it makes sense to set cx.current instead, and remember to unset it when you're done? That would make the diff a lot smaller and also reduce the number of parameters everywhere. You could add a with_current_path function or something to make sure you don't forget to reset it.

src/test/rustdoc/cross-crate-primitive-doc.rs

src/librustdoc/html/format.rs

src/librustdoc/html/tests.rs

src/librustdoc/html/render/mod.rs

jyn514 · 2021-04-16T17:33:52Z

Oh definitely - this PR is fine as-is, I meant in a follow-up.

… r=jyn514 rustdoc: clean up and test macro visibility print This fixes the overly-complex invariant mentioned in <rust-lang#83237 (comment)>, where the macro source can't have any links in it only because the cache hasn't been populated yet.

jyn514 · 2021-04-16T20:17:15Z

I opened #84258 to get #84074 and #84246 merged; please wait for that to land (it should hopefully be the last rebase you have to do). I'll be offline the rest of the day and I didn't want you to hit more conflicts overnight :)

Thanks so much for working on this!

@bors delegate=notriddle

bors · 2021-04-16T20:17:17Z

✌️ @notriddle can now approve this pull request

…=jyn514 rustdoc: clean up and test macro visibility print This fixes the overly-complex invariant mentioned in <rust-lang#83237 (comment)>, where the macro source can't have any links in it only because the cache hasn't been populated yet.

Instead of using a depth counter and adding "../" to get to the top, this commit makes rustdoc actually compare the path of what it's linking from to the path that it's linking to. This makes the resulting HTML shorter. Here's a comparison of one of the largest (non-source) files in the Rust standard library docs (about 4% improvement before gzipping). $ wc -c struct.Wrapping.old.html struct.Wrapping.new.html 2387389 struct.Wrapping.old.html 2298538 struct.Wrapping.new.html Most if it can be efficiently gzipped away. $ wc -c struct.Wrapping.old.html.gz struct.Wrapping.new.html.gz 70679 struct.Wrapping.old.html.gz 70050 struct.Wrapping.new.html.gz But it also makes a difference in the final DOM size, reducing it from 91MiB to 82MiB.

The function `to_src_with_space` doesn't even accept the cache as a parameter, so it doesn't matter any more what's in it.

This should help make things consistent.

notriddle · 2021-04-17T18:39:49Z

@bors r=jyn514

bors · 2021-04-17T18:39:50Z

📌 Commit 755b4fb has been approved by jyn514

Rollup of 4 pull requests Successful merges: - rust-lang#83237 (rustdoc: use more precise relative URLs) - rust-lang#84150 (rustdoc: move some search code into search.js) - rust-lang#84203 (rustdoc: Give a more accurate span for anchor failures) - rust-lang#84257 (Add documentation to help people find `Ipv4Addr::UNSPECIFIED`) Failed merges: r? `@ghost` `@rustbot` modify labels: rollup

jyn514 · 2021-04-18T03:48:23Z

🎉 🎉 🎉

…mp-to-definition, r=GuillaumeGomez rustdoc: use more precise URLs for jump-to-definition links As an example, this cuts down <https://doc.rust-lang.org/nightly/nightly-rustc/src/rustc_middle/ty/mod.rs.html> by about 11%. $ du -h new_mod.rs.html old_mod.rs.html 296K new_mod.rs.html 332K old_mod.rs.html Like rust-lang#83237, but separate code since source links have a different URL structure. Related to [Zulip discussion](https://rust-lang.zulipchat.com/#narrow/stream/266220-rustdoc/topic/RFC.20for.20.22jump.20to.20definition.22.20feature/near/299029786) and [the jump-to-definition pre-RFC](GuillaumeGomez/rfcs#1).

…o-decl, r=notriddle rustdoc: Treat declarative macros more like other item kinds Apparently at some time in the past we were unable to generate an href for the module path inside the visibility of decl macros 2.0 (`pub(in ...)`). As a result of this, a whole separate function was introduced specifically for printing the visibility of decl macros that didn't attempt to generate any links. The description of PR rust-lang#84074 states: > This fixes the overly-complex invariant mentioned in rust-lang#83237 (comment), where the macro source can't have any links in it only because the cache hasn't been populated yet. I can no longer reproduce the original issue. Reusing the existing visibility rendering logic *seems* to work just fine (I couldn't come up with any counterexamples, though I invite you to prove me wrong). * Fixes rust-lang#83000 * Fixes the visibility showing up "twice" in rustdoc-JSON output: Once as the `visibility` field, once baked into the source[^1] * Fixes `#[doc(hidden)]` not getting rendered on doc(hidden) decl macros 2.0 under `--document-hiden-items` (for decl macros 1.2 the issue remains; I will address this separately when fixing rust-lang#132304). --- <details><summary>Outdated Section</summary> NOTE: The current version of this PR is committing a UI crime, I'd like to receive feedback on that. Maybe you have a satisfactory solution for how to remedy it. Namely, as you know we have two different ways of / modes for highlighting code with color: 1. Only highlighting links / item paths and avoiding to highlight tokens by kind like keywords (to reduce visual noise and maybe also artifact size). Used for item declarations(\*). 2. Highlighting tokens by kind. Used for code blocks written by the user. (\*): With the notable exception being macro declarations! Well, since this PR reuses the same function for rendering the item visibility (which only makes sense), we have a clash of modes: We now use both ways of highlighting code for decl macros: №1 for the visibility, №2 for the rest. This awkward. See for yourself: * On master: ![Screenshot 2024-10-29 at 03-37-48 by_example_vis_named in decl_macro a b c - Rust](https://github.com/user-attachments/assets/22f0ab6e-9ba9-4c4e-8fb0-0741c91d360b) * On this branch: ![Screenshot 2024-10-29 at 03-36-41 by_example_vis_named in decl_macro a b c - Rust](https://github.com/user-attachments/assets/b11d81a3-3e2e-43cb-a5b8-6773a3048732) </details> Furthermore, we now no longer syntax-highlight declarative macros (be it `macro_rules!` or `macro`) since that was inconsistent with the way we render all other item kinds. See (collapsed) *Outdated Section* above. See also rust-lang#132302 (comment). | On master | On this branch | |---|---| | ![Screenshot 2024-11-13 at 16-12-46 by_example_vis_named in decl_macro a b c - Rust](https://github.com/user-attachments/assets/cb3aeb42-a56d-4ced-80d9-f2694f369af1) | ![Screenshot 2024-11-13 at 16-13-22 by_example_vis_named in decl_macro a b c - Rust](https://github.com/user-attachments/assets/b73bee50-1b85-4862-afba-5ad471443ccc) | [^1]: E.g., `"visibility":{"restricted":{"parent":1,"path":"::a"}},/*OMITTED*/,"inner":{"macro":"pub(in a) macro by_example_vis_named($foo:expr) {\n ...\n}"}`

Rollup merge of rust-lang#132302 - fmease:rustdoc-better-vis-for-macro-decl, r=notriddle rustdoc: Treat declarative macros more like other item kinds Apparently at some time in the past we were unable to generate an href for the module path inside the visibility of decl macros 2.0 (`pub(in ...)`). As a result of this, a whole separate function was introduced specifically for printing the visibility of decl macros that didn't attempt to generate any links. The description of PR rust-lang#84074 states: > This fixes the overly-complex invariant mentioned in rust-lang#83237 (comment), where the macro source can't have any links in it only because the cache hasn't been populated yet. I can no longer reproduce the original issue. Reusing the existing visibility rendering logic *seems* to work just fine (I couldn't come up with any counterexamples, though I invite you to prove me wrong). * Fixes rust-lang#83000 * Fixes the visibility showing up "twice" in rustdoc-JSON output: Once as the `visibility` field, once baked into the source[^1] * Fixes `#[doc(hidden)]` not getting rendered on doc(hidden) decl macros 2.0 under `--document-hiden-items` (for decl macros 1.2 the issue remains; I will address this separately when fixing rust-lang#132304). --- <details><summary>Outdated Section</summary> NOTE: The current version of this PR is committing a UI crime, I'd like to receive feedback on that. Maybe you have a satisfactory solution for how to remedy it. Namely, as you know we have two different ways of / modes for highlighting code with color: 1. Only highlighting links / item paths and avoiding to highlight tokens by kind like keywords (to reduce visual noise and maybe also artifact size). Used for item declarations(\*). 2. Highlighting tokens by kind. Used for code blocks written by the user. (\*): With the notable exception being macro declarations! Well, since this PR reuses the same function for rendering the item visibility (which only makes sense), we have a clash of modes: We now use both ways of highlighting code for decl macros: №1 for the visibility, №2 for the rest. This awkward. See for yourself: * On master: ![Screenshot 2024-10-29 at 03-37-48 by_example_vis_named in decl_macro a b c - Rust](https://github.com/user-attachments/assets/22f0ab6e-9ba9-4c4e-8fb0-0741c91d360b) * On this branch: ![Screenshot 2024-10-29 at 03-36-41 by_example_vis_named in decl_macro a b c - Rust](https://github.com/user-attachments/assets/b11d81a3-3e2e-43cb-a5b8-6773a3048732) </details> Furthermore, we now no longer syntax-highlight declarative macros (be it `macro_rules!` or `macro`) since that was inconsistent with the way we render all other item kinds. See (collapsed) *Outdated Section* above. See also rust-lang#132302 (comment). | On master | On this branch | |---|---| | ![Screenshot 2024-11-13 at 16-12-46 by_example_vis_named in decl_macro a b c - Rust](https://github.com/user-attachments/assets/cb3aeb42-a56d-4ced-80d9-f2694f369af1) | ![Screenshot 2024-11-13 at 16-13-22 by_example_vis_named in decl_macro a b c - Rust](https://github.com/user-attachments/assets/b73bee50-1b85-4862-afba-5ad471443ccc) | [^1]: E.g., `"visibility":{"restricted":{"parent":1,"path":"::a"}},/*OMITTED*/,"inner":{"macro":"pub(in a) macro by_example_vis_named($foo:expr) {\n ...\n}"}`

rust-highfive assigned jyn514 Mar 17, 2021

rust-highfive added the S-waiting-on-review Status: Awaiting review from the assignee but also interested parties. label Mar 17, 2021