support regular expression ?? about jmespath.jep

[notjames]

I don't think regular expressions should be part of the core JMESPath syntax because different programming languages have different regular expression implementations, so it wouldn't be possible to guarantee that a regular expression run using JMESPath on Python would have the exact same results on PHP or Ruby or Lua or Go.

Instead I suggest writing your own custom function that makes Python regular expressions available to your code: https://github.com/jmespath/jmespath.py/blob/0.10.0/README.rst#custom-functions

I completely disagree with this assertion. Though many languages do indeed have different implementations, I estimate 99.999% of them use a PCRE library/implementation for regexes as PCRE has become the defacto standard for regular expressions. Moreover, if every language used the rationale to defer to implement regular expressions because other languages do it differently then no language would implement them. I find myself deferring to jq 90% of the time because jp doesn't employ regular expressions.

[simonw]

jq is an interesting comparison here. It actually bundles the full implementation of the regular expression library (a submodule that imports Oniguruma: https://github.com/stedolan/jq/tree/master/modules) that it uses in the C implementation of the tool - then different programming languages have bindings that depend on that shared implementation: https://github.com/stedolan/jq/wiki/FAQ#language-bindings

jmespath works differently: rather than shipping a single C implementation that is shared by different languages, it encourages entirely fresh implementations in different languages with no shared code, but sharing instead a thorough set of compliance tests: https://jmespath.org/libraries.html

Tests are here: https://github.com/jmespath/jmespath.test/tree/master/tests

This approach could be used to add regular expression support, by spinning up a thorough compliance test and maybe defining a subset of regular expression syntax that must be supported in order for an implementation to pass the test.

It would be a lot of work though!

[Harliff]

+1 vote for regex support!

[simonw]

Here's a really interesting proposal, for an interoperable subset of regular expressions: https://cabo.github.io/iregexp/draft-bormann-jsonpath-iregexp.html

It's still in early stages, but it looks like it's directly relevant to this conversation.

[mswezey23]

I'd be in huge support for adding regex support!

[simonw]

I don't think regular expressions should be part of the core JMESPath syntax because different programming languages have different regular expression implementations, so it wouldn't be possible to guarantee that a regular expression run using JMESPath on Python would have the exact same results on PHP or Ruby or Lua or Go.

Instead I suggest writing your own custom function that makes Python regular expressions available to your code: https://github.com/jmespath/jmespath.py/blob/0.10.0/README.rst#custom-functions

[simonw]

There's a useful comparison table of the differences between different language implementations here: https://web.archive.org/web/20130830063653/http://www.regular-expressions.info:80/refflavors.html