(Jump to navigation)

Python RegEx through the ages, Appendix B: Other third-party regular expression modules

← Return to main document

B.1. The Wahlig PCRE binding

PyPI package: python-pcre

Arkadiusz Wahlig’s third-party (second) pcre module binding interfaces with the current version of the PCRE 1 library (not PCRE 2; it uses pcre.h, not pcre2.h). It accordingly supports the syntax supported by the current version of PCRE 1. For details of this syntax, its readme refers to the documentation of PCRE’s PHP binding.

B.1.1. API of the second pcre module

Despite the module name, it does not match the API of the older, low-level module by the same name (it actually uses its own low-level _pcre support module). Rather, its API is substantially the same as re with a handful of important differences:

• By default, the substitution syntax matches str.format, i.e. sub and friends behave like the second regex module’s subf and friends by default.
• The global enable_re_template_mode() function changes this so that the substitution syntax works like re. This recalls the syntax flag system from the first regex module in that it is done globally, but it furthermore offers no straightforward way of undoing it (technically, you can undo it with the decidedly kludgy pcre.Match = pcre.REMatch.__bases__[0]).
• Its readme describes it as lacking the “scanner APIs” of the re module. I am not sure what this refers to.
• The available flags are somewhat different, but with a large overlap.

B.1.2. Flags of the second pcre module

Below is the list of flags as they currently appear in PCRE 1, along with the names under which they are exported in the Wahlig binding, and also their values. I call your attention to the following:

• The LOCALE flag is absent, as with (less surprisingly) the DEBUG flag and (utterly unsurprisingly) the TEMPLATE flag.
• The VERBOSE, ANCHORED, MULTILINE and DOTALL flags have different numerical values in the current PCRE 1 library than in the old and custom version used by the pre module.
• A few of the flags which were supported but not exported by the pre (and first pcre) modules are now exported.
• Many, many more flags are supported by the current version of PCRE 1, only a few of which are exported, though more could theoretically be used as magic numbers.
• Because the PCRE_NEWLINE_* flags are not supposed to coëxist with each other, they are assigned permutations of two of the bits, rather than being assigned individual bits. Consequently, PCRE_NEWLINE_CRLF and PCRE_NEWLINE_ANYCRLF are not powers of two.

B.1.3. The _pcre C support module

TODO