Minimisation should not combine end states with distinct sets of endids. #461
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This was adapted from the group capture integration branch. It isn't specific to group capture. Currently, metadata on end states will not prohibit merging, which can lead to false positives when using endids to detect which of the original regexes are matching when several regexes are compiled and then unioned into a single DFA. This adds a pass before the main minimisation loop that splits the ECs of end states with distinct sets of end IDs, which is sufficient to prevent them from merging later. I reworked the test tests/endids/endids2_union_many_endids.c somewhat, since it was previously checking that the end states WERE merged. Unfortunately the other checks are somewhat weakened -- I removed the part checking that `endids[j] == j+1` because that ordering isn't guaranteed anymore: the patterns aren't anchored, so some examples will match more than one of them, and the endids collected can be nonconsecutive. I added a separate test, endids10_minimise_partial_overlap.c, which checks the minimal case that when the two regexes /^abc$/ and /^ab*c$/ are combined and a distinct endid is set on each, matching "ac" "abc" "abbc" gets one or both endids as appropriate. Ideally, we'd have a property test that checks that any arbitrary set of regex patterns has the same set of inputs matching -> endids when they are run individually vs. when combined into a single FSM, determinised, and minimised.
These got missed. Also, rename EXPENSIVE_INTEGRITY_CHECKS in minimise.c to EXPENSIVE_CHECKS and move the <fsm/print.h> header inclusion.
|
I fixed two things caught by the CI run:
But I still need to update the minimization test oracle to prevent it from merging end states with distinct sets of endids. |
If two end states have distinct sets of endids associated with them, mark them as distinguishable before entering the main fixpoint loop. This brings the results from the test oracle back in line with fsm_minimise.
katef
approved these changes
Apr 26, 2024
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
This was adapted from the group capture integration branch. It isn't specific to group capture.
Currently, metadata on end states will not prohibit merging, which can lead to false positives when using endids to detect which of the original regexes are matching when several regexes are compiled and then unioned into a single DFA. This adds a pass before the main minimisation loop that splits the ECs of end states with distinct sets of end IDs, which is sufficient to prevent them from merging later.
I reworked the test tests/endids/endids2_union_many_endids.c somewhat, since it was previously checking that the end states WERE merged. Unfortunately the other checks are somewhat weakened -- I removed the part checking that
endids[j] == j+1because that ordering isn't guaranteed anymore: the patterns aren't anchored, so some examples will match more than one of them, and the endids collected can be nonconsecutive.I added a separate test, endids10_minimise_partial_overlap.c, which checks the minimal case that when the two regexes /^abc$/ and /^ab*c$/ are combined and a distinct endid is set on each, matching "ac" "abc" "abbc" gets one or both endids as appropriate. Ideally, we'd have a property test that checks that any arbitrary set of regex patterns has the same set of inputs matching -> endids when they are run individually vs. when combined into a single FSM, determinised, and minimised.