dafact

Encodes data as ASP facts.

dafact solves the reiterative task of encoding a dataset into a set of Answer Set Programming facts. The resulting ASP program, which can be partially personalized, can be encoded as python clingo objects directly or as plain text, and it handles floating point numbers automatically.

Data can be fed into dafact through CSV files, numpy arrays, pandas DataFrames and other typical formats.

It can be used both as a python library and as a command line tool.

Installation

The tool is easily installable through pip:

python3 -m pip install dafact

Short usage

A more detailed usage guide can be found here.

Python

Dafacter python class provides all the funcionality. Once fed with data, a Dafacter object provides two main methods for obtaining a logic program:

from dafact import Dafacter
dafacter = Dafacter("data/haberman.csv", have_names=True) # Fed data into the object
clingo_facts = dafacter.as_clingo_facts()  # Returns a list of clingo.Function objects
program_text = dafacter.as_program_string()  # Returns the program as plain text

The following piece of code loads the haberman dataset from a csv file and encodes it as a logic program.

# examples/usage_csv.py
from dafact import Dafacter

dafacter = Dafacter("data/haberman.csv", have_names=True)
print(dafacter.as_program_string())

The result of that code would be:

feature("age"). feature("op_year"). feature("nodes"). feature("survival").
instance(0). value(0,"age",30). value(0,"op_year",64). value(0,"nodes",1). value(0,"survival",1).
instance(1). value(1,"age",30). value(1,"op_year",62). value(1,"nodes",3). value(1,"survival",1).
(. . .)

The encoded style can also be tweaked easily, and it accepts different kind of data sources from numpy arrays to pandas DataFrames. A more detailed guide on usage of the python library can be found in examples folder.

Command line tool

Once installed through pip users can use dafact for directly obtain a logic program from csv files through the use of the command line tool. The usage of the tool is the same to the use of the Dafacter python class for csv files.

~/$ dafact --help
usage: dafact [-h] [--feature-names [FEATURE_NAMES [FEATURE_NAMES ...]]] [--factor FACTOR]
              [--numerical-columns [NUMERICAL_COLUMNS [NUMERICAL_COLUMNS ...]]] [--have-names] [--omit-names] [--delimiter DELIMITER]
              infile outfile

Dafact CLI Encodes data as ASP facts.

positional arguments:
  infile                Input csv file.
  outfile               Ouput ASP program.

optional arguments:
  -h, --help            show this help message and exit

Options:
  --feature-names [FEATURE_NAMES [FEATURE_NAMES ...]]
                        Feature names for the csv columns.
  --factor FACTOR       factor help
  --numerical-columns [NUMERICAL_COLUMNS [NUMERICAL_COLUMNS ...]]
                        Indexes for numerical columns.
  --have-names          Must be if csv have the name of the columns in the first line.
  --omit-names          Used together with --have-names for omitting the names in the file.
  --delimiter DELIMITER
                        Field delimiter for the csv file.