Fast parser for fixed-column line data files

pip install lineparser==0.0.0.dev7



lineparser is a small library with one goal: parse fixed-width formatted files extremely quickly. In order to achieve this, lineparser uses Cython to obtain the speed offered by good old C code and the convenience of Python.

Documentation can be found here.


You should be able to install lineparser via pip if you are on a Windows or Linux 64 bit machine.

$ pip3 install lineparser


import lineparser
import time

fields = [(lineparser.Float64, 12), (lineparser.Float64, 10), (lineparser.Float64, 12), 
        (lineparser.String, 6), (lineparser.String, 6), (lineparser.String, 14),
        (lineparser.String, 14), (lineparser.Float64, 6)]

    start = time.time()
    # On a successfull parse, result will be a list of lists and numpy arrays (strings will be in 
    # lists, and numbers will be in numpy arrays). 
    result = lineparser.parse(fields, b'data/small_data.par')
    end = time.time()
    print(f"Took {end - start} seconds to parse")
except lineparser.LineParsingError as e:
    print(f"Encountered the following error while trying to parse:\n {str(e)}")


    31.43339 6.531E-28   31.442390     3     2       0.00048       0.00000   100
    41.89467 1.415E-26   62.878170     4     3       0.00065       0.00000   100
    41.89786 3.538E-27   62.876840     4     3       0.00064       0.00000   100

Running the example: python3

Installing from Source

Installing from source is also easy. You must have GCC installed on your machine, and you must have Cython installed only if you want to modify the library. Then run this command:

$ python3 install


lineparser is simple to build, and should only require one command:

$ python3 build_ext --inplace

You should then be able to import it in a python interpreter (in the build directory):

$ python3 build_ext --inplace
$ python3 
Python 3.6.7 (default, Oct 22 2018, 11:32:17)
[GCC 8.2.0] on linux
Type "help", "copyright", "credits" or "license" for more information.
>>> import lineparser
>>> ...