Conversation
rustbot
added
S-waiting-on-review
|
Implementation looks good to me. Let's confirm the perf gain. Also I'd like @lolbinarycat to have a look as she worked on this recently (iirc). @bors try @rust-timer queue |
This comment has been minimized.
This comment has been minimized.
rustbot
added
the
S-waiting-on-perf
This comment has been minimized.
This comment has been minimized.
fix quadratic naming of duplicate sidebar links
This comment has been minimized.
This comment has been minimized.
| @@ -0,0 +1,21 @@ | |||
| // Regression test for <https://github.com/rust-lang/rust/issues/158174>. | |||
There was a problem hiding this comment.
I feel like this should be a rustc-perf benchmark, not a regular test, since this is a performance improvement, not a bugfix. @GuillaumeGomez what do you think?
There was a problem hiding this comment.
Why not both? =D
But yes, very good idea!
| struct UsedLinks { | ||
| links: FxHashSet<String>, | ||
| /// Next number to try for each anchor. | ||
| next_suffix: FxHashMap<String, usize>, | ||
| } | ||
|
|
||
| fn get_next_url(used_links: &mut UsedLinks, url: String) -> String { | ||
| if used_links.links.insert(url.clone()) { | ||
| return url; | ||
| } | ||
| let mut add = 1; | ||
| while !used_links.insert(format!("{url}-{add}")) { | ||
| add += 1; | ||
| let add = used_links.next_suffix.entry(url.clone()).or_insert(1); | ||
| loop { | ||
| let candidate = format!("{url}-{add}"); | ||
| *add += 1; | ||
| if used_links.links.insert(candidate.clone()) { | ||
| return candidate; | ||
| } | ||
| } |
There was a problem hiding this comment.
Unless I'm missing an edge case somehow, this should be doable with 1 hashmap and no loop.
| struct UsedLinks { | |
| links: FxHashSet<String>, | |
| /// Next number to try for each anchor. | |
| next_suffix: FxHashMap<String, usize>, | |
| } | |
| fn get_next_url(used_links: &mut UsedLinks, url: String) -> String { | |
| if used_links.links.insert(url.clone()) { | |
| return url; | |
| } | |
| let mut add = 1; | |
| while !used_links.insert(format!("{url}-{add}")) { | |
| add += 1; | |
| let add = used_links.next_suffix.entry(url.clone()).or_insert(1); | |
| loop { | |
| let candidate = format!("{url}-{add}"); | |
| *add += 1; | |
| if used_links.links.insert(candidate.clone()) { | |
| return candidate; | |
| } | |
| } | |
| type UsedLinks = FxHashMap<String, usize>; | |
| fn get_next_url(used_links: &mut UsedLinks, url: String) -> String { | |
| let count = used_links.entry(url.clone()).or_insert(0); | |
| let res = if count == 0 { url } else { format!("{url}-{count}") }; | |
| *count += 1; | |
| res | |
| } |
There was a problem hiding this comment.
Yes, a counter is enough and better indeed . Will make the change with a bit of correctness here ( *count == 0) as in the suggested snippet it compares a &mut usize with a number . Which won't compile . Thank you for the review!
|
Finished benchmarking commit (cc18f18): comparison URL. Overall result: no relevant changes - no action neededBenchmarking means the PR may be perf-sensitive. Consider adding rollup=never if this change is not fit for rolling up. @rustbot label: -S-waiting-on-perf -perf-regression Instruction countThis perf run didn't have relevant results for this metric. Max RSS (memory usage)Results (primary 3.8%, secondary -3.7%)A less reliable metric. May be of interest, but not used to determine the overall result above.
CyclesResults (secondary -3.2%)A less reliable metric. May be of interest, but not used to determine the overall result above.
Binary sizeResults (primary -0.1%, secondary -0.1%)A less reliable metric. May be of interest, but not used to determine the overall result above.
Bootstrap: 498.333s -> 499.277s (0.19%) |
rustbot
removed
the
S-waiting-on-perf
|
Hum... Do we even have a test to stress test the sidebar? Should we add the new sidebar perf test? |
it doesn't seem like it
yes, that would be good, maybe even add a larger test using the python script in the original issue. |
|
If the perf. regression can be shown with a simple crate, I think it would be fine to add it as a stress test benchmark to |
|
Thanks! I'll open a follow-up rustc-perf PR with a stress benchmark based on the script from #158174 . |
|
Please add the new perf test first, like that we can have a more precise view of the impact of this PR (in addition to the numbers you already provided). |
6 participants
This PR fixes quadratic naming of duplicate sidebar links as discussed in #158174.
Fixes #158174
get_next_urlgives duplicate sidebar links with unique names but it started counting from 1 every time which makes the page quadratic.So to make it more optimal I changed that to remember the next unique number for each name making it linear per page.
Also added the test at
tests/rustdoc-html/deref/sidebar-links-deref-chain.rsResults
r? @GuillaumeGomez