Is it possible to have a task-specific spelling.txt that can be supplied as a parameter to the CLI?

[ralienpp]

I am trying to put together a prototype that leverages LanguageTool to check the grammar and style of strings and comments in source code files (Python, in my example).

There is a tool that preprocesses the source code and extracts all relevant strings into a separate file, called result.txt. This file is then fed into LanguageTool like so java -jar languagetool-commandline.jar -l en-US --line-by-line --disable WHITESPACE_RULE result.txt.

Generally, it works, but there are a lot of complaints that could have been avoided, if I could explicitly tell LanguageTool which words are correct as they are. For example, consider this output:

2.) Line 187, column 90, Rule ID: MORFOLOGIK_RULE_EN_US premium: false prio=-10
Message: Possible spelling mistake found.
Suggestion: bad Request
...ed names representing failure info, such as `badRequest`.         Defaults to `None`.         - `tex...
                                                ^^^^^^^^^^

3.) Line 189, column 61, Rule ID: MORFOLOGIK_RULE_EN_US premium: false prio=-10
Message: Possible spelling mistake found.
...s`: A list of or a single text used for the `PKIStatusText` structure. Defaults to         `"This text ...
                                                ^^^^^^^^^^^^^

In the business logic of my program, the terms "badRequest" and "PKIStatusText" are correct.

I've read that I can address this problem by updating my spelling.txt in org\languagetool\resource\en\hunspell. Although it does the trick, my concern is that I am making a global change. The terms in question are fine for program X, but they're not applicable to another program I work on. Ideally, there ought to be a way to pass the path to my custom spelling.txt as an argument to languagetool-commandline.jar, such that I can apply different exceptions to different projects.

How can this be accomplished? Or perhaps there is a better alternative that I am unaware of?

[danielnaber]

I don't think there's such a feature. Your best bet is to think of LT as a library and use its output as the input for your own filtering layer that removes these matches.