Custom parser

[tmarshall]

If using a specific logger in python, it is useful to set a custom parser for data printed by python. I've set this as a parser option, which is optional, outside of the normal mode. This means you can leave mode as json, for example, and send json to the python script. Then, when receiving lines from the script, it will first check for parser and then fall back to modes.

For example, if you have a script that outputs something like this:
2015-05-13 13:32:52,812 INFO dosomething:12 root.main() - xyz abc
You can set a parser:

var options = {
  mode: 'json',
  parser: function(line) {
    var expr = /^([\d-]+ [\d:]+),\d* (\w+) (\w+):(\d+) ([\w\.\(\)]+) - (.*)$/;
    var results = expr.exec(line);

    if (!results) {
      throw 'Failed to parse python output';
    }

    return {
      time: new Date(results[1]),
      type: results[2].toLowerCase(),
      module: results[3],
      lineno: +results[4],
      func: results[5],
      content: results[6]
    };
  }
};

This will return

{ time: Wed May 13 2015 13:32:52 GMT-0700 (PDT),
  type: 'info',
  module: 'dosomething',
  lineno: 12,
  func: 'root.main()',
  content: 'xyz abc' }

If a line is invalid it will throw, and python-shell will return a proper error.

[extrabacon]

Parsing with a function sure is useful. I would go further and eliminate the mode parameter in order to use functions only. I would still use strings as shortcuts for "text" and "json", but use the new function-based parser parameter instead.

I will make some changes in a new branch in order to test this.

[tmarshall]

👍

[extrabacon]

Take a look at the changes in https://github.com/extrabacon/python-shell/tree/parsers.

[tmarshall]

ah I like the formatter as well. lgtm