LightCSV

Simple light CSV reader

This CSV reader is implemented in just pure Python. It allows to specify a separator, a quote char and column titles (or get the first row as titles). Nothing more, nothing else.

Installation

Install it with:

pip install lightcsv

Usage

Usage is pretty straightforward:

from lightcsv import LightCSV

for row in LightCSV().read_file("myfile.csv"):
    print(row)

This will open a file named myfile.csv and iterate over the CSV file returning each row as a key-value dictionary. Line endings can be either \n or \r\n. The file will be opened in text-mode with utf-8 encoding.

You can supply your own stream (i.e. an open file instead of a filename). You can use this, for example, to open a file with a different encoding, etc.:

from lightcsv import LightCSV

with open("myfile.csv") as f:
    for row in LightCSV().read(f):
        print(row)

NOTE: Blank lines at any point in the file will be ignored

Parameters

LightCSV can be parametrized during initialization to fine-tune its behaviour.

The following example shows initialization with default parameters:

from lightcsv import LightCSV

myCSV_reader = LightCSV(
    separator=",",
    quote_char='"',
    field_names = None,
    strict=True,
    has_headers=False
)

Available settings:

• separator: character used as separator (defaults to ,)
• quote_char: character used to quote strings (defaults to ").
  This char can be escaped by duplicating it.
• field_names: can be any iterable or sequence of str (i.e. a list of strings).
  If set, these will be used as column titles (dictionary keys), and also sets the expected number of columns.
  
• strict: Sets whether the parser runs in strict mode or not.
  In strict mode the parser will raise a ValueError exception if a cell cannot be decoded or column numbers don't match. In non-strict mode non-recognized cells will be returned as strings. If there are more columns than expected they will be ignored. If there are less, the dictionary will contain also fewer values.
• has_headers: whether the first row should be taken as column titles or not.
  If set, field_names cannot be specified. If not set, and no field names are specified, dictionary keys will be just the column positions of the cells.

Data types recognized

The parser will try to match the following types are recognized in this order:

• None (empty values). Unlike CSV reader, it will return None (null) for empty values.
  Empty strings ("") are recognized correctly.
• str (strings): Anything that is quoted with the quotechar. Default quotechar is ".
  If the string contains a quote, it must be escaped duplicating it. i.e. "HELLO ""WORLD""" decodes to HELLO "WORLD" string.
• int (integers): an integer with a preceding optional sign.
• float: any float recognized by Python
• datetime: a datetime in ISO format (with 'T' or whitespace in the middle), like 2022-02-02 22:02:02
• date: a date in ISO format, like 2022-02-02
• time: a time in ISO format, like 22:02:02

If all this parsing attempts fails, a string will be returned, unless strict_mode is set to True. In the latter case, a ValueError exception will be raised.

Implementing your own type recognizer

You can implement your own deserialization by subclassing LightCSV and override the method parse_obj().

For example, suppose we want to recognize hexadecimal integers in the format 0xNNN.... We can implement it this way:

import re
from lightcsv import LightCSV

RE_HEXA = re.compile('0[xX][A-Za-z0-9]+$')  # matches 0xNNNN (hexadecimals)


class CSVHexRecognizer(LightCSV):
    def parse_obj(self, lineno: int, chunk: str):
        if RE_HEXA.match(chunk):
            return int(chunk[2:], 16)
        
        return super().parse_obj(lineno, chunk)

As you can see, you have to override parse_obj(). If your match fails, you have to invoke super() (overridden) parse_obj() method and return its result.

---

Why

Python built-in CSV module is a bit over-engineered for simple tasks, and one normally doesn't need all bells and whistles. With LightCSV you just open a filename and iterate over its rows.

Decoding None for empty cells is needed very often and can be really cumbersome as the standard csv tries hard to cover many corner-cases (if that's your case, this tool might not be suitable for you).