Optimize header parsing - #513
Conversation
bc1c4f8 to
5fa05e5
Compare
Fixes gettalong#505 Benchmarks: "setext": ```bash ruby -rbenchmark -Ilib -rkramdown -e 'p Benchmark.measure{Kramdown::Document.new("1#{" "*20000}2\n==\n")}' ``` "atx": ```bash ruby -rbenchmark -Ilib -rkramdown -e 'p Benchmark.measure{Kramdown::Document.new("## 1#{" "*20000}2")}' ```
krasnoukhov
left a comment
krasnoukhov
left a comment
There was a problem hiding this comment.
Seems like this fixes a performance issue with no breaking changes. Good work!
gettalong
left a comment
gettalong
left a comment
There was a problem hiding this comment.
Thanks for the pull request and your great work!
Before I merge, please also remove the changes to .gitignore and remove the zero-byte test/testcases/block/04_header/atx_header.hcd file.
| @src.check(ATX_HEADER_MATCH) | ||
| level, text, id = @src[1], @src[2].to_s.strip, @src[3] | ||
| text, id = parse_header_contents | ||
| text.sub!(/[\t ]#+\z/, '') && text.rstrip! |
There was a problem hiding this comment.
Is there a reason why these two statements are on one line? I would rather prefer them on separate lines.
There was a problem hiding this comment.
There is no need to call rstrip! if sub! return nil.
The other options here are:
A:
text.rstrip! if text.sub!(...)
B:
if text.sub!(...)
text.rstrip!
end
I find the current one (&&) to be the most readable but can go with A or B if that's what you prefer.
There was a problem hiding this comment.
Personally I would just put them on separate lines and take this minimal performance hit because it would be easier to see what's going on without the additional conditional. But I'm okay with leaving this as it is.
|
|
||
| HEADER_ID=/(?:[ \t]+\{#([A-Za-z][\w:-]*)\})?/ | ||
| SETEXT_HEADER_START = /^(#{OPT_SPACE}[^ \t].*?)#{HEADER_ID}[ \t]*?\n(-|=)+\s*?\n/ | ||
| SETEXT_HEADER_START = /^#{OPT_SPACE}(?<contents>.*)\n(?<level>[-=])[-=]*[ \t\r\f\v]*\n/ |
There was a problem hiding this comment.
Why did you leave out [^ \t] in the new version? It is not strictly necessary for the main kramdown parser due to the fixed order of block parsers but behaviour of derived parser may change. As far as I can tell, it doesn't make a performance difference.
Also: What is the reason behind changing \s to [ \t\r\f\v]?
There was a problem hiding this comment.
Why did you leave out [^ \t] in the new version?
Because it's not necessary as we have to validate header length later on anyway. I have now added it back in. No performance difference.
Also: What is the reason behind changing \s to [ \t\r\f\v]?
I find constructs like \s*?\n a bit cryptic. Performance-wise both are the same.
By the way, do we actually want to accept \r\f\v here? CommonMark spec only mentions trailing spaces (https://spec.commonmark.org/0.28/#example-54), and also allows OPT_SPACE for the underline.
|
|
||
| HEADER_ID = /[\t ]{#(?<id>[A-Za-z][\w:-]*)}\z/ | ||
|
|
||
| # @return [[String, String]] header text and optional ID. |
There was a problem hiding this comment.
Please use standard RDoc comments.
There was a problem hiding this comment.
I'm not familiar with RDoc, and although I've read the RDoc Markup Reference, it doesn't seem to mention the correct markup for type hints, so I'm not sure what to do here.
The format I've used is YARD, which is also widely supported by IDEs for type hints. Can you convert to the desired format when merging?
There was a problem hiding this comment.
RDoc doesn't have type hints.
Just leave out this API documentation (most of the other parsing methods don't have API documentation as well).
|
|
||
| # @param [Number] level | ||
| # @param [String] text | ||
| # @param [String, nil] id |
There was a problem hiding this comment.
Please use standard RDoc comments.
There was a problem hiding this comment.
Same here, just remove the API documentation.
|
I just noticed that the PR doesn't include the test cases themselves. Could you add them or should I add them after I merge the changes? And please squash the commits into one before I merge them - thank you! |
Please add them yourself, it will be faster this way.
Nowadays you can do it yourself at merge time by clicking "Squash and merge"! |
|
Merged - thanks! |
|
@gettalong Will you cut a release please? Thank you |
|
Yes, if I have time this weekend. |
3 participants
Fixes #505
Benchmarks:
"setext":
ruby -rbenchmark -Ilib -rkramdown -e 'p Benchmark.measure{Kramdown::Document.new("1#{" "*20000}2\n==\n")}'"atx":
ruby -rbenchmark -Ilib -rkramdown -e 'p Benchmark.measure{Kramdown::Document.new("## 1#{" "*20000}2")}'