Repository navigation
Add string predicates startsWith, endsWith and contains #78
Description
What do you think about adding predicates to the CodeQL type string for determining whether a string has a given prefix or suffix or contains a substring? For example startsWith(string), endsWith(string) and contains(string).
Currently the workarounds are using indexOf(...) = 0 or matches(...%) (which seems to be faster than indexOf, see github/codeql#6479 (comment)). However, these predicates do not convey the intention as clearly, might not be that performant and for matches one must take care not to accidentally use %or _ where the intention was to match them literally.
In the github/codeql repository (at github/codeql@3953331) there are at least:
- 196 cases where
startsWithcould be used
(I searched for the regexmatches\("[^%_]*%"\)in CodeQL source files) - 72 cases where
endsWithcould be used
(I searched for the regexmatches\("%[^%_]*"\)in CodeQL source files)
Activity
adityasharad commentedon Aug 23, 2021
Thanks for the suggestion!
To give you some insight about what happens under the hood in the CodeQL engine,
matches("...%")gets translated into an operation likestartsWith("..."), and similarlymatches("%...")becomesendsWith("..."). So the engine is already doing that optimisation for you. Usingmatchesin this way is indeed preferable overindexOf, becauseindexOfhas to produce all(character, index)pairs.This isn't high on our list of priorities, but I agree that either exposing
startsWith/endsWithprimitives, or updating the docs formatchesto clearly describe this case, is a good way to convey the intention.[-]Add `string` predicates `startsWith` and `endsWith`[/-][+]Add `string` predicates `startsWith`, `endsWith` and `contains`[/+]Marcono1234 commentedon Sep 7, 2021
A
contains(string)predicate would be useful as well, have added it to the description.