Currently regex are used to match values against a pattern, but error messages quality is poor.

When pattern constraint is not used abusively (i.e. when it would have been OK to use a number), it's be possible to use a PEG grammar parser library or a parser library to produce better error messages.

Technical steps

• create a custom check (named like PegCheck) taking a grammar and a column name (or many)
• for each field using a pattern constraint:
  • don't use the pattern constraint in the schema, just use "type": "string"
  • explain in the field description the form of expected value
  • use examples (even if not yet supported, cf https://github.com/frictionlessdata/specs/issues/384)
  • configure the custom check for this field
• define grammars outside, in .peg files, and reference them by path (?)

Resources

Parsers:

• https://github.com/pyparsing/pyparsing

PEG grammars parsers:

• https://github.com/erikrose/parsimonious
• https://github.com/igordejanovic/Arpeggio
• article: https://tomassetti.me/parsing-in-python/

Other approaches:

• https://reparse.readthedocs.io/en/latest/