Skip to content

Does a TRIE-EXACT only pattern do unnecessary work? #23864

@richardleach

Description

@richardleach

Description
If you run: perl -e 'use re q|debug|; my $x = "x" x 10; 1 while $x =~ /xx|ab/aag;',
the pattern has compiled to a TRIE-EXACT node:

Compiling REx "xx|ab"
Final program:
   1: TRIE-EXACT[ax] (9)
      <xx> 
      <ab> 
   9: END (0)

The output for each match is something like:

stclass AHOCORASICK-EXACT[ax] minlen 2 
Matching REx "xx|ab" against "xxxxxxxxxx"
Matching stclass AHOCORASICK-EXACT[ax] against "xxxxxxxxxx" (10 bytes)
   0 <> <xxxxxxxxxx>         |   0|  Charid:  1 CP:  78 State:    1, word=0 - legal
   1 <x> <xxxxxxxxx>         |   0|  Charid:  1 CP:  78 State:    2, word=0 - legal
   2 <xx> <xxxxxxxx>         |   0|  Charid:  1 CP:  78 State:    3, word=1 - accepting
Matches word #1 at position 0. Trying full pattern...
   0 <> <xxxxxxxxxx>         |   0| 1:TRIE-EXACT[ax](9)
   0 <> <xxxxxxxxxx>         |   0| TRIE: State:    1 Accepted: N TRIE: Charid:  1 CP:  78 After State:    2
   1 <x> <xxxxxxxxx>         |   0| TRIE: State:    2 Accepted: N TRIE: Charid:  1 CP:  78 After State:    3
   2 <xx> <xxxxxxxx>         |   0| TRIE: State:    3 Accepted: Y TRIE: Charid:  0 CP:   0 After State:    0
                             |   0| TRIE: got 1 possible matches
                             |   0| TRIE matched word #1, continuing
                             |   0| TRIE: only one match left, short-circuiting: #1 <xx>
   2 <xx> <xxxxxxxx>         |   0| 9:END(0)
Match successful!

Is it necessary to try the full TRIE-EXACT pattern when in effect, AHOCORASICK-EXACT[ax] has already found a whole pattern match?

Perl configuration
blead

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Type

    No type

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions