Using original text during sampling

When sampling possible strings from a template like (A | B), hassil would previously produce: a and b (normalized). It will now produce the original strings with whitespace: A and B.

No skipping whitespace

Bugfix to not allow skipping whitespace while parsing under certain conditions.

Version 1.0.0

Remove slow ANTLR parser and use a custom one (~8-10x faster).

Breaking Changes

• Whitespace is now significant around the | character. Previously, light(s | ing) would match both "lights" and "lighting", but now it will match "lights" and "light ing"
• Only punctuation is skipped automatically, instead of all "non-word" characters (regex \W). Templates must now include % and ° explicitly. The current set of punctuation characters is .。,，?¿？!！;；:：
• Template parsing is more strict, so unmatched ( or ] characters will cause parsing errors

New Features

Intents YAML has a new settings block:

language: "..."
settings:
  ignore_whitespace: false

The only option is ignore_whitespace, which is intended for language families like zh (Chinese) that may be written without spaces. The default value is False.

Other Changes

• The ANTLR runtime is no longer needed as a dependency

Correct whitespace in sampler

v0.1.3

Merge branch 'main' of https://github.com/home-assistant/hassil

Initial release

v0.1.2

Add CI workflow