Cleanup cudf::strings::detail::regex_parser class source #10975
Conversation
davidwendt
added
2 - In Progress
cpp/src/strings/regex/regcomp.cpp
Outdated
| } d; | ||
| }; | ||
| std::vector<Item> m_items; | ||
| std::vector<regex_parser::Item> expand_counted_items() |
There was a problem hiding this comment.
This is not a new function but was simply moved from the regex_compiler class to here since it only needs the regex_parser::_items variable.
Codecov Report
@@ Coverage Diff @@
## branch-22.08 #10975 +/- ##
===============================================
Coverage ? 86.34%
===============================================
Files ? 144
Lines ? 22738
Branches ? 0
===============================================
Hits ? 19632
Misses ? 3106
Partials ? 0 Continue to review full report at Codecov.
|
davidwendt
added
3 - Ready for Review
cpp/src/strings/regex/regcomp.cpp
Outdated
| int quoted = nextc(yy); | ||
| _chr = 0; | ||
| char32_t chr = 0; | ||
| auto quoted = next_char(chr); |
There was a problem hiding this comment.
Don't know about performance but if the signature was cuda::std::pair<bool> next_char() then this would look slightly neater auto [is_quoted, chr] = next_char()
cpp/src/strings/regex/regcomp.cpp
Outdated
| int32_t _id_ccls_w{-1}; // alphanumeric | ||
| int32_t _id_ccls_W{-1}; // not alphanumeric | ||
| int32_t _id_ccls_s{-1}; // space | ||
| int32_t _id_ccls_d{-1}; // digit | ||
| int32_t _id_ccls_D{-1}; // not digit |
There was a problem hiding this comment.
At the risk of making the code too verbose, can these get vowels and more elaborate docs?
|
@gpucibot merge |
This cleans up the awkward range literals for supporting the `CCLASS` and `NCCLASS` regex instructions. The range values were always paired (first,last) but arranged consecutively in a flat vector so `[idx] and [idx+1]` were range pairs `idx` was even. This PR introduces a `reclass_range` class that holds the pairs so we can use normal algorithms to manipulate them. There is some overlap with code changes in PR #10975 Reference #3582 Authors: - David Wendt (https://github.com/davidwendt) Approvers: - Mike Wilson (https://github.com/hyperbolic2346) - MithunR (https://github.com/mythrocks) URL: #11045
Cleans up the
regex_parserclass source to fix member names, comments, and simplify some logic creating parse-items.Minimal changes were made to the
regex_compilerto accommodate some public interface changes.Also, the
regex_compiler::expand_counted()function was moved intoregex_parsersince it only need the parser class'_itemsdata. It seemed more apt for the member function to part ofregex_parserthanregex_compiler.Reference #3582