26 lines
1.1 KiB
Text
26 lines
1.1 KiB
Text
|
The .words file is used by gospel (v1.2+), which wraps the Hunspell libraries
|
||
|
|
but populates the dictionary with identifiers from the Go source.
|
||
|
|
|
||
|
|
<https://github.com/kortschak/gospel>
|
||
|
|
|
||
|
|
Alas, no comments are allowed in the .words file and newer versions of gospel
|
||
|
|
error out on seeing them. This is really a hunspell restriction.
|
||
|
|
|
||
|
|
We assume en_US hunspell dictionaries are installed and used.
|
||
|
|
The /AFFIXRULES are defined in en_US.aff (eg: /usr/share/hunspell/en_US.aff)
|
||
|
|
Invoke `hunspell -D` to see the actual locations.
|
||
|
|
|
||
|
|
Words which are in the base dictionary can't have extra affix rules added to
|
||
|
|
them, so we have to start with the affixed variant we want to add.
|
||
|
|
Thus `creds` rather than `cred/S` and so on.
|
||
|
|
|
||
|
|
So we can't use receive/DRSZGBU, adding 'U', to allow unreceive and variants,
|
||
|
|
we have to use unreceive as the stem.
|
||
|
|
|
||
|
|
We can't define our own affix or compound rules,
|
||
|
|
to capture rfc\d{3,} or 0x[0-9A-Fa-f]{2}
|
||
|
|
|
||
|
|
The spelling tokenizer doesn't take "permessage-deflate" as allowing for ...
|
||
|
|
"permessage-deflate", which is an RFC7692 registered extension for websockets.
|
||
|
|
We have to explicitly list "permessage".
|