semi structured xml to dict

ssxtd is an xmlreader similar to xmltodict, but supporting semi structured xml, and providing a more flexible environnment. ssxtd use either of :

• the native xml package
• the lxml package
• the defusedxml package

Globally, stick to native xml package. It's the faster for most situations.

Note : ssxtd was created to parse very big files, as a result, the default parsing depth is 2 and the parsing functions are generators.

Quickstart

pip install ssxtd

from ssxtd import parsers
result = next(parsers.xml_parse(my_file, depth=0))

What if ...

...i want to parse big files?

use :
parsers.xml_iterparse(my_file)

note : by default, it will return depth 2 elements.

...my xml file has mixed tag and text?

ssxtd will automatically convert mixed tags and text to a string, keeping the right order.

...i want to use defusedxml to secure my app?

use :
parsers.dxml_parse(my_file)
or
parsers.dxml_iterparse(my_file)

...i want to use lxml?

use :
parsers.lxml_parse(my_file)
or
parsers.lxml_iterparse(my_file)

note : it will be slower than xml_parse and xml_iterparse

Options

depth

you can adjust the depth level of the returned objects, even when not using iterparse.

note : you can't use depth = 0 when using iterparse

trim_spaces

will trim spaces for each value found. can be usefull when you have some ugly xml like:

<root>
  <text>we have
  some indentation
  problems
  </text>
</root>

del_empty (default to True)

if set to False, will not remove empty tags

cleanup_namespaces

if set to False, will not remove namespaces

verbose

if set to True, will show a progression bar \o/

recover

if set to True, will recover from malformed xml ( cf test_malformed_xml.py)

note : lxml_parse and lxml_iterparse will use the lxml abilities whereas the others will use a BeautifulSoup transformation

compression

ssxtd can manage ZIP, GZIP, ByteIO, and path to files

for ZIP and GZIP, you must set the parameter "compression" to either "gz" or "zip"

parsers.xml_parse(my_file, compression="gz"):

you can also set the parameter to "auto", ssxtd will then auto detect the file type from the extension (.xml, .zip, or .gz)

note : atm, in a zip compression mode, only .xml files situated at the root of the zip file will be read

object_processorr

if you specify the parameter "object_processor=my_function" when calling a parser, your function will be called for each object

see bin/run_exemple.py

Allows to do special actions like merging tags directly during the parsing

value_processor

if you specify the parameter "value_processor=my_function" when calling a parser, your function will be called for each value found

e.g a simple type conversion :

def try_conversion(value):
        try:
            return int(value)
        except (ValueError, TypeError):
            pass
        try:
            return float(value)
        except (ValueError, TypeError):
            pass
        return value

Requirements

Python

python >= v3.7.0b1
due to https://github.com/python/cpython/commit/066df4fd454d6ff9be66e80b2a65995b10af174f
you CAN use older version of pythons ( i tested up to 3.5) but you won't be able to read zip files

Libs

• bs4
• tqdm

Run tests

install pytest:
pip install pytest
in the root directory, run :
pytest
for running a single file, place yourself at the root folder and run :
python -c "import ssxtd.tests.test_malformed_xml"

Performances of the parsing functions

time to parse https://ftp.ncbi.nlm.nih.gov/pubmed/baseline/pubmed19n0001.xml.gz
GZ file size : 19MB
extracted size : 185MB
proc : i7-7700HQ

Function	XML file	GZ file	ZIP file (no compression)
xml_parse	32.76501545000065	36.07715339999959	33.419777400000385
xml_iterparse	37.56028480000168	42.16279835000023	39.137448499999664
lxml_parse	37.464776250000796	38.880011499999455	37.046347550000064
lxml_iterparse	47.04024449999997	45.959421049999946	45.05521540000154
dxml_parse	41.52063830000043	40.07632935000038	38.88691465000011
dxml_iterparse	45.195273199999065	44.895784000000276	44.13825424999959

for much more details please see ssxtd\benchmarks\results\result.csv