sten0 / gospell Goto Github PK

******************************** WORK IN PROGRESS ****************************
******************************************************************************

This project attempts to automate spell checking for comments in Go code. Most
modern languages rely on comments for documentation, and I was curious to see
to what degree this can be automated. It's split into a few packages:

1) Package "check" contains generic logic for spell checking given an alphabet
   and dictionary. It works as a standalone spell checking package.
2) Package "lang" store dictionary data and currently only supports
   English(US).
3) Package "scrape" has logic to scrape data from Merriam-Webster's online
   dictionary.
4) Package "main" builds a binary to run predefined or tunable spell checkers
   against specific files or recursively on a directory. Java, C, C++, and
   Scala files are also supported, with the default being Go.

So what's considered a misspelling?

Since comments are often code expressions and not valid grammar, it's not
rational to simply check each space-delimited string against a dictionary.
The default behavior classifies a misspelled word if it:

    - has at least 5 characters
        and
            - differs by 1 character insertion
            or
            - differs by 1 character deletion
            or
            - differs by a single consecutive character swap

To run the classifier on a Go project:

    $ ./gospell .

There's minimal tuning support. To classify against words that:

    - are at least 4 characters long
    and
    - differ by at most 2 insertions:

Try:

    $ ./gospell -ml=4 -mi=2 .

This is a decent start. Results often need pruning by a human eye. It may be
worth exploring the following features:

      -- the frequency of a misspelled word wherein some threshold
         declassifies the misspelling
      -- adapt the insertion, deletion, and swap restrictions
         based on the size of the word,
         so longer words can differ by more changes.