rebl

Welcome to rebl - a Regular Expression Based Linter.

Linters are thoroughly useful tools in a bid to improve code quality. However typically their behaviour is fixed or they are difficult to extend.

While one can use grep to scan a file for a pattern, when scanning a file for many patterns this gets either slow (because the file is scanned once per pattern) or unmaintainable (as the patterns get very, very long indeed).

rebl is a small linter which seeks to balance the simplicity of grep with the usefulness of a regular linter.

Aren't regular expressions unsuitable to deal with language syntax?

Absolutely. This linter was developed with the full understanding that line-based regular expressions have such limitations. This linter solidly intends to provide 80% usability with 20% of the effort.

Stuff you need to know

The pattern configuration file for this linter is a python module. Why should you care? For two reasons -

• the file lives in its own directory, and
• that directory should also contain an empty __init__.py file.

The default pattern configuration file is kept in (path of rebl)/.reblrc/config.py but additional paths will be searched (see below).

There are two ways to override the default path; one is to specify --config=/absolute/path/to/config.py on the command line. The other is to have a file called .reblrc containing --config=relative/path/to/config.py. Note that if a file .reblrc isn't found in the current working directory, rebl will look in the parent directory. If no .reblrc file is found anywhere it proceeds checking further paths. The relative/path/to/config.py is relative to the directory where the .reblrc file was found.

If no pattern configuration is specified, rebl will look for one

• in the current working directory, then checking parent directories (project- or multi-project config - likely to hit user's home directory)
• in the rebl program directory (user-wide config)
• in /etc/rebl/config.py (i.e. system wide config)

Example pattern configuration

For your convenience an example pattern configuration is provided in the .reblrc directory of this repository. A symbolic link _reblrc is provided to make it visible.

Example config:

patterns = {}
patterns['.py'] = {
  "HW0025": (
    "No need for exc_info when using log.exception",
    ["log.exception", "exc_info"],[],[] ),
}

where

• .py is the file extension to which the pattern applies

• HW0025 is pattern key, must be unique for each pattern Fist char is confidence (HML for High, Medium, Low) Second char is error level (EWI for Error, Warning, Info) Next set of digits is the rest of the pattern unique identifier

• String given on the next line is the user friendly linter message, 1 line

• next line has 3 lists, the "all of", "any of" and "none of" lists.

  A line flags up when

  • it matches all of the regexes in the "all of" list
  • it matches any of the regexes in the "any of" list
  • it matches none of the regexes in the "none of" list.

  Regexes are automatically anchored.

  This means that a pattern hello will match any line containing the word "hello" - it is automatically rewritten to "^.*hello.*$" to make rebl behave more like grep.

Advanced detection

All of, any of and none of lists can be combined.

If that's not enough, If all three lists are empty, rebl will look for a function called detect_[ext]_[pattern key] which accepts filename and line arguments - for example,

def detect_py_HW0025(filename, line):
    return True if 'hello' in line else False

will be called for each line and should

• return True on match; or
• return False on no match.

Fixers

If a function fix_[ext]_[pattern key] exists, this can be used to perform fixes in batch. for example,

def fix_py_MW0049(filename, line):
    return line.replace("import csv", "import unicodecsv as csv")

would automatically be invoked if rebl is run with --fix and lines match the pattern as given for pattern key MW0049.

Line hook/context collector

It is possible to define a line hook to collect context on every line, e.g.

def linehook_<ext>(filename, lines, linenum, context):

This will be called on every line if defined - permits collecting state data such as current function name, errors found thusfar, whatever.

As the linehook is called frequently, ideally it should be kept light; that is, try to avoid too many loops in here.

State should be kept in dict context - this dict is reset each file.