regular expressions

[mflatt]

This regular expression sublanguage is based on @CooperCorad's BS thesis.

Rendered documentation: https://users.cs.utah.edu/~mflatt/tmp/rhombus-rx/rhombus/regexp.html

Different from Racket:

• The rx sublanguage is encoded in shrubbery notation, not a string. There's also a subsublanguage for character sets (i.e., the subsublanguage within []).

• Matching is whole-input by default. To select partial-input matching, use rx_in to construct a regexp or use a method like RX.match_in instead of RX.match.

• The primary syntax for a capture group gives it a name, instead of just a position. These names are bound as variables when rx or rx_in is used as a binding form, and they serve as symbol keys in a map (within an object) produced by methods like RX.match.

• The rx form is not provided by just #lang racket. Instead, use import rhombus/rx open, which introduces only a few bindings in the expression and binding spaces but many bindings in the regexp-pattern and character-class spaces.

Inherited from Racket:

• A regexp matches either at the byte or character level. Both kinds can be used on bytes, byte strings, and input ports.

This draft includes some changes relative to the version shown in at the August 22 meeting (#180) based on feedback in that meeting — so, thanks to Robby, Alex, and Ben.

The documentation has lots of examples, and you find more in the test suite: https://github.com/mflatt/rhombus/blob/rx/rhombus/rhombus/tests/rx.rhm

[mflatt]

Updated:

• removed ^ and end-of-line $, added bol and eol
• changed any pattern to be any character, but . is still any except newline
• removed enable_newline and disable_newline
• changed lookback to lookbehind (which seems to be the more established term)
• removed ^ as charset complement, keeping ! as complement

[mflatt]

There's certainly more to do for improving documentation (maybe after splitting the Rhombus manual into guide and reference documents), but I'll merge, and we can take further improvements from there.