Fork

This version of pyspellcode has been hacked up to be used on the Stan Math library. Many of the improvements are not Stan specific, consider forking off the non_stan_specific_changes branch.

pyspellcode

Python script for using clang and hunspell for spell checking source code comments. It's a little hacky but it "works for me". pyspellcode implements two strategies for extracting comments:

1. (default) Use Clang's built-in AST dump tool. This will traverse all included files, including third party libraries (this is inefficient), but it will only spellcheck in the specified files. It will make some attempt to sensibly-parse doxygen comments and avoid "spellchecking" LaTeX fragments, etc. It will only spellcheck comments that are attached to declarations, in particular, it will not catch this typo:

// This is attached to the declaration of foo.
void foo() {
  // This typooo filled comment is not attached to a declaration.
}

2. Use a specialized Clang tool, build on the LibTooling API. This will traverse only the indicated files, skipping includes (much faster), and it will extract all comments in the specified files. It is necessary to build the tool first, which takes a while. It is likely that updates to clang will break the build process. It will report all your LaTeX fragments as unrecognized words.

Usage

0. (install Clang and hunspell)
1. git clone https://github.com/peterwicksstringfield/pyspellcode/tree/non_stan_specific_changes
2. ./spell-check.py example/example1.cpp ...
3. cd some/where/else
4. path/to/spell-check.py some_file.cpp
5. (optionally ...)
6. cd path/to/spell-check.py
7. (install ninja, the build system the llvm project uses)
8. ./spell-check.py --build-tool
9. (wait for a long time)
10. ./spell-check.py --use-tool example/example1.cpp ...
11. cd some/where/else
12. path/to/spell-check.py --use-tool --path-to-tool path/to some_file.cpp

For the most up-to-date command line argument usage, run the script with the --help flag (-h for short). For example:

usage: spell-check.py [-h] [-v] [-I <dir>] [-std=c99] [-std=c++11]
                      [-std=c++14] [-std=c++17] [--doxygen-only] [-e]
                      [--show-file-progress] [-p <full-file-path>] [-c]
                      [-x <extra-argument-to-clang>] [--use-tool]
                      [--build-tool] [--path-to-tool <path-to-tool>]
                      [filename [filename ...]]

Extract and spellcheck comments from provided C++ source.

positional arguments:
  filename              filename to inspect

optional arguments:
  -h, --help            show this help message and exit
  -v, --verbose         gets more verbose to aid with diagnostics
  -I <dir>, --include-dir <dir>
                        adds directory to include search path, ignored if
                        --use-tool is set
  -std=c99              selects the C99 language standard
  -std=c++11            selects the C++11 language standard (default)
  -std=c++14            selects the C++14 language standard
  -std=c++17            selects the C++17 language standard
  --doxygen-only        skip normal comments (// and /* */), only check
                        doxygen comments (/// and /** */)
  -e, -Werror, --error-exit
                        nonzero exit status for unrecognized words
  --show-file-progress  shows filenames and results even when no unrecognized
                        words
  -p <full-file-path>, --personal-dict <full-file-path>
                        specify the fullpath to a personal dictionary
  -c, --collect         output deduplicated list of unrecognized words
  -x <extra-argument-to-clang>, --extra-clang-arg <extra-argument-to-clang>
                        extra argument for clang
  --use-tool            use specialized Clang Tool to extract comments
  --build-tool          build specialized Clang Tool; slow!
  --path-to-tool <path-to-tool>
                        path to specialized build tool, default to CWD

Additional

Thanks to Louis Langholtz for the original implementation. https://github.com/louis-langholtz/pyspellcode

and to Daniel Beard for the Clang Tool code. https://gist.github.com/daniel-beard?page=1

and to the developers of Clang and hunspell!