XmlXdiff 0.3.2

Compare xml files with svg output.

XmlXdiff

XmlXdiff was inspired by X-Diff.

Since version 0.3.2 the distance cost's algorithm is replaced by parent-identification. This might by a wrong decision but the result's for huge xml documents (see. test 9) improved in performance and quality.

This is not a bullet prove library (till now). It s more a playground to get in touch with comparing tree structures and presenting the resulting in a charming way.

dependencies

• PySide2
• svgwrite
• lxml

installation

python pip XmlXdiff

fist step

from XmlXdiff.XReport import DrawXmlDiff

_xml1 = """<root><deleted>with content</deleted><unchanged/><changed name="test1" /></root>"""
_xml2 = """<root><unchanged/><changed name="test2" /><added/></root>"""

with open("test1.xml", "w") as f:
    f.write(_xml1)

with open("test2.xml", "w") as f:
    f.write(_xml2)

x = DrawXmlDiff("test1.xml", "test2.xml")
x.saveSvg('xdiff.svg')

status quo

implementation

Each xml element is identified by it's xpath and a hash calculated by selecting relevant information. Start with the identification of huge xml blocks (changed/moved). Identification of parent elements by tag, text-pre, text-post, attribute-names and attribute-values. Parent xml blocks can contain further parent xml blocks.

 <tag attribute-name:"attribute-value" ...> 
 text-pre 
 	<... children ...>
 text-post
 </tag>

1. mark all xml elements as changed
2. iterate over parent blocks, starting with maximum children to parent blocks with less children
3. mark unchanged xml elements of current parent
4. mark moved xml elements of current parent
5. mark xml elements identified by tag name and attribute names of the current parent
6. mark xml elements identified by attributes values and element text of the current parent
7. mark xml elements identified by tag name of the current parent
8. mark xml elements with xpath that do not exist in the other xml tree as added/deleted of the current parent
9. Repeat 3. till all xml elements are identified

All xml elements that are still marked as changed have to be investigated

performance

test1: delta_t=0.0625s xml_elements=63
test2: delta_t=0.0156s xml_elements=5
test3: delta_t=0.0156s xml_elements=4
test4: delta_t=0.0313s xml_elements=32
test5: delta_t=0.0312s xml_elements=34
test6: delta_t=0.0156s xml_elements=34
test7: delta_t=0.0156s xml_elements=8
test8: delta_t=0.0937s xml_elements=67
test9: delta_t=5.3894s xml_elements=6144
test11: delta_t=0.0292s xml_elements=34
test12: delta_t=0.0312s xml_elements=45
test13: delta_t=0.0625s xml_elements=75

coverage

pyscript   NoSource: No source for code: 'C:\Portable\git\XmlXdiff\pyscript'.
Aborting report output, consider using -i.
Name                               Stmts   Miss  Cover
------------------------------------------------------
lib\XmlXdiff\XDiffer.py              169     24    86%
lib\XmlXdiff\XHash.py                 88      3    97%
lib\XmlXdiff\XPath.py                 57      5    91%
lib\XmlXdiff\XReport\XRender.py       60     26    57%
lib\XmlXdiff\XReport\__init__.py     329     49    85%
lib\XmlXdiff\XTypes.py               145     40    72%
lib\XmlXdiff\__init__.py               3      0   100%
------------------------------------------------------
TOTAL                                851    147    83%

open issues

• performance analysis and improvements (different hash algorithms, ...)
• if there are some users, improve interface
• investigation of merge interfaces

release notes

v0.3.2:

• implemented parent-identification without children context
• split segments replaced by parent-identification (no dependency to number of child's nor content of child's)
• color scheme changed
• coverage improved

v0.2.2:

• search areas are split into segments between unchanged xml nodes
• added/deleted/verified to be added
• overlapping search areas possible now (merge proposals)

documentation