Handle regexes with unicode escape sequences in .find and .findAll

[maxbrunsfeld]

Fixes atom/atom#16126

Background

The TextBuffer.find family of methods use PCRE to perform regex matching directly on the buffer's native contents. This has a number of advantages: it eliminates a lot of string copying, removes to need for conversions between raw character indices and row/column coordinates, and allows searching to be done on a background thread if desired. See #5, #35.

Problem

There are some differences between PCRE's and ECMAScript's regex syntax. One difference is that PCRE does not support the \u00df syntax for specifying UTF16 character codes.

Solution

Luckily, it's trivial to convert these character sequences into their actual UTF16 values. I've added that conversion in this PR.

Future Work

There may be other incompatibilities between PCRE and ECMAScript regexes. One other example I know of is the unicode code point escape sequence, which are slightly different than the UTF16 character code escape sequences I've dealt with here. They should be easy to support as well, though I have not done so here.

/cc @t9md

[nathansobo]

💖

[leroix]

It would be nice if there was a standalone c++ module version of v8's Irregexp.

[nathansobo]

It would be nice if there was a standalone c++ module version of v8's Irregexp.

I have lamented this as well.

[maxbrunsfeld]

Yeah, we looked into that a little bit, but concluded that Irregexp is totally coupled to V8's internals - I think it generates native code using the same JIT machinery that V8 uses for compiling JS code.

Luckily, PCRE is a pretty lightweight dependency.