====================
Benchmarks and Speed
====================
:Author:
Stefan Behnel
.. meta::
:description: Performance evaluation of lxml and ElementTree:
fast operations, common pitfalls and optimisation hints.
:keywords: Python XML parser performance, XML processing, performance comparison,
lxml performance, lxml.etree, lxml.objectify, benchmarks, ElementTree
lxml.etree is a very fast XML library. Most of this is due to the
speed of libxml2, e.g. the parser and serialiser, or the XPath engine.
Other areas of lxml were specifically written for high performance in
high-level operations, such as the tree iterators.
On the other hand, the simplicity of lxml sometimes hides internal
operations that are more costly than the API suggests. If you are not
aware of these cases, lxml may not always perform as you expect. A
common example in the Python world is the Python list type. New users
often expect it to be a linked list, while it actually is implemented
as an array, which results in a completely different complexity for
common operations.
Similarly, the tree model of libxml2 is more complex than what lxml's
ElementTree API projects into Python space, so some operations may
show unexpected performance. Rest assured that most lxml users will
not notice this in real life, as lxml is very fast in absolute
numbers. It is definitely fast enough for most applications, so lxml
is probably somewhere between 'fast enough' and 'the best choice' for
yours. Read some messages_ from happy_ users_ to see what we mean.
.. _messages: http://permalink.gmane.org/gmane.comp.python.lxml.devel/3250
.. _happy: http://article.gmane.org/gmane.comp.python.lxml.devel/3246
.. _users: http://thread.gmane.org/gmane.comp.python.lxml.devel/3244/focus=3244
This text describes where lxml.etree (abbreviated to 'lxe') excels, gives
hints on some performance traps and compares the overall performance to the
original ElementTree_ (ET) and cElementTree_ (cET) libraries by Fredrik Lundh.
The cElementTree library is a fast C-implementation of the original
ElementTree.
.. _ElementTree: http://effbot.org/zone/element-index.htm
.. _cElementTree: http://effbot.org/zone/celementtree.htm
.. contents::
..
1 How to read the timings
2 Bad things first
3 Parsing and Serialising
4 The ElementTree API
5 Tree traversal
6 XPath
7 lxml.objectify
General notes
=============
First thing to say: there *is* an overhead involved in having a DOM-like C
library mimic the ElementTree API. As opposed to ElementTree, lxml has to
generate Python representations of tree nodes on the fly when asked for them,
and the internal tree structure of libxml2 results in a higher maintenance
overhead than the simpler top-down structure of ElementTree. What this means
is: the more of your code runs in Python, the less you can benefit from the
speed of lxml and libxml2. Note, however, that this is true for most
performance critical Python applications. No one would implement Fourier
transformations in pure Python when you can use NumPy.
The up side then is that lxml provides powerful tools like tree iterators,
XPath and XSLT, that can handle complex operations at the speed of C. Their
pythonic API in lxml makes them so flexible that most applications can easily
benefit from them.
How to read the timings
=======================
The statements made here are backed by the (micro-)benchmark scripts
`bench_etree.py`_, `bench_xpath.py`_ and `bench_objectify.py`_ that come with
the lxml source distribution. They are distributed under the same BSD license
as lxml itself, and the lxml project would like to promote them as a general
benchmarking suite for all ElementTree implementations. New benchmarks are
very easy to add as tiny test methods, so if you write a performance test for
a specific part of the API yourself, please consider sending it to the lxml
mailing list.
The timings presented below compare lxml 3.1.1 (with libxml2 2.9.0) to the
latest released versions of ElementTree (with cElementTree as accelerator
module) in the standard library of CPython 3.3.0. They were run
single-threaded on a 2.9GHz 64bit double core Intel i7 machine under
Ubuntu Linux 12.10 (Quantal). The C libraries were compiled with the
same platform specific optimisation flags. The Python interpreter was
also manually compiled for the platform. Note that many of the following
ElementTree timings are therefore better than what a normal Python
installation with the standard library (c)ElementTree modules would yield.
Note also that CPython 2.7 and 3.2+ come with a newer ElementTree version,
so older Python installations will not perform as good for (c)ElementTree,
and sometimes substantially worse.
.. _`bench_etree.py`: https://github.com/lxml/lxml/blob/master/benchmark/bench_etree.py
.. _`bench_xpath.py`: https://github.com/lxml/lxml/blob/master/benchmark/bench_xpath.py
.. _`bench_objectify.py`: https://github.com/lxml/lxml/blob/master/benchmark/bench_objectify.py
The scripts run a number of simple tests on the different libraries, using
different XML tree configurations: different tree sizes (T1-4), with or
without attributes (-/A), with or without ASCII string or unicode text
(-/S/U), and either against a tree or its serialised XML form (T/X). In the
result extracts cited below, T1 refers to a 3-level tree with many children at
the third level, T2 is swapped around to have many children below the root
element, T3 is a deep tree with few children at each level and T4 is a small
tree, slightly broader than deep. If repetition is involved, this usually
means running the benchmark in a loop over all children of the tree root,
otherwise, the operation is run on the root node (C/R).
As an example, the character code ``(SATR T1)`` states that the benchmark was
running for tree T1, with plain string text (S) and attributes (A). It was
run against the root element (R) in the tree structure of the data (T).
Note that very small operations are repeated in integer loops to make them
measurable. It is therefore not always possible to compare the absolute
timings of, say, a single access benchmark (which usually loops) and a 'get
all in one step' benchmark, which already takes enough time to be measurable
and is therefore measured as is. An example is the index access to a single
child, which cannot be compared to the timings for ``getchildren()``. Take a
look at the concrete benchmarks in the scripts to understand how the numbers
compare.
Parsing and Serialising
=======================
Serialisation is an area where lxml excels. The reason is that it
executes entirely at the C level, without any interaction with Python
code. The results are rather impressive, especially for UTF-8, which
is native to libxml2. While 20 to 40 times faster than (c)ElementTree
1.2 (which was part of the standard library before Python 2.7/3.2),
lxml is still more than 10 times as fast as the much improved
ElementTree 1.3 in recent Python versions::
lxe: tostring_utf16 (S-TR T1) 7.9958 msec/pass
cET: tostring_utf16 (S-TR T1) 83.1358 msec/pass
lxe: tostring_utf16 (UATR T1) 8.3222 msec/pass
cET: tostring_utf16 (UATR T1) 84.4688 msec/pass
lxe: tostring_utf16 (S-TR T2) 8.2297 msec/pass
cET: tostring_utf16 (S-TR T2) 87.3415 msec/pass
lxe: tostring_utf8 (S-TR T2) 6.5677 msec/pass
cET: tostring_utf8 (S-TR T2) 76.2064 msec/pass
lxe: tostring_utf8 (U-TR T3) 1.1952 msec/pass
cET: tostring_utf8 (U-TR T3) 22.0058 msec/pass
The difference is somewhat smaller for plain text serialisation::
lxe: tostring_text_ascii (S-TR T1) 2.7738 msec/pass
cET: tostring_text_ascii (S-TR T1) 4.7629 msec/pass
lxe: tostring_text_ascii (S-TR T3) 0.8273 msec/pass
cET: tostring_text_ascii (S-TR T3) 1.5273 msec/pass
lxe: tostring_text_utf16 (S-TR T1) 2.7659 msec/pass
cET: tostring_text_utf16 (S-TR T1) 10.5038 msec/pass
lxe: tostring_text_utf16 (U-TR T1) 2.8017 msec/pass
cET: tostring_text_utf16 (U-TR T1) 10.5207 msec/pass
The ``tostring()`` function also supports serialisation to a Python
unicode string object, which is currently faster in ElementTree
under CPython 3.3::
lxe: tostring_text_unicode (S-TR T1) 2.6896 msec/pass
cET: tostring_text_unicode (S-TR T1) 1.0056 msec/pass
lxe: tostring_text_unicode (U-TR T1) 2.7366 msec/pass
cET: tostring_text_unicode (U-TR T1) 1.0154 msec/pass
lxe: tostring_text_unicode (S-TR T3) 0.7997 msec/pass
cET: tostring_text_unicode (S-TR T3) 0.3154 msec/pass
lxe: tostring_text_unicode (U-TR T4) 0.0048 msec/pass
cET: tostring_text_unicode (U-TR T4) 0.0160 msec/pass
For parsing, lxml.etree and cElementTree compete for the medal.
Depending on the input, either of the two can be faster. The (c)ET
libraries use a very thin layer on top of the expat parser, which is
known to be very fast. Here are some timings from the benchmarking
suite::
lxe: parse_bytesIO (SAXR T1) 13.0246 msec/pass
cET: parse_bytesIO (SAXR T1) 8.2929 msec/pass
lxe: parse_bytesIO (S-XR T3) 1.3542 msec/pass
cET: parse_bytesIO (S-XR T3) 2.4023 msec/pass
lxe: parse_bytesIO (UAXR T3) 7.5610 msec/pass
cET: parse_bytesIO (UAXR T3) 11.2455 msec/pass
And another couple of timings `from a benchmark`_ that Fredrik Lundh
`used to promote cElementTree`_, comparing a number of different
parsers. First, parsing a 274KB XML file containing Shakespeare's
Hamlet::
xml.etree.ElementTree.parse done in 0.017 seconds
xml.etree.cElementTree.parse done in 0.007 seconds
xml.etree.cElementTree.XMLParser.feed(): 6636 nodes read in 0.007 seconds
lxml.etree.parse done in 0.003 seconds
drop_whitespace.parse done in 0.003 seconds
lxml.etree.XMLParser.feed(): 6636 nodes read in 0.004 seconds
minidom tree read in 0.080 seconds
And a 3.4MB XML file containing the Old Testament::
xml.etree.ElementTree.parse done in 0.038 seconds
xml.etree.cElementTree.parse done in 0.030 seconds
xml.etree.cElementTree.XMLParser.feed(): 25317 nodes read in 0.030 seconds
lxml.etree.parse done in 0.016 seconds
drop_whitespace.parse done in 0.015 seconds
lxml.etree.XMLParser.feed(): 25317 nodes read in 0.022 seconds
minidom tree read in 0.288 seconds
.. _`from a benchmark`: http://svn.effbot.org/public/elementtree-1.3/benchmark.py
.. _`used to promote cElementTree`: http://effbot.org/zone/celementtree.htm#benchmarks
Here are the same benchmarks again, but including the memory usage
of the process in KB before and after parsing (using os.fork() to
make sure we start from a clean state each time). For the 274KB
hamlet.xml file::
Memory usage: 7284
xml.etree.ElementTree.parse done in 0.017 seconds
Memory usage: 9432 (+2148)
xml.etree.cElementTree.parse done in 0.007 seconds
Memory usage: 9432 (+2152)
xml.etree.cElementTree.XMLParser.feed(): 6636 nodes read in 0.007 seconds
Memory usage: 9448 (+2164)
lxml.etree.parse done in 0.003 seconds
Memory usage: 11032 (+3748)
drop_whitespace.parse done in 0.003 seconds
Memory usage: 10224 (+2940)
lxml.etree.XMLParser.feed(): 6636 nodes read in 0.004 seconds
Memory usage: 11804 (+4520)
minidom tree read in 0.080 seconds
Memory usage: 12324 (+5040)
And for the 3.4MB Old Testament XML file::
Memory usage: 10420
xml.etree.ElementTree.parse done in 0.038 seconds
Memory usage: 20660 (+10240)
xml.etree.cElementTree.parse done in 0.030 seconds
Memory usage: 20660 (+10240)
xml.etree.cElementTree.XMLParser.feed(): 25317 nodes read in 0.030 seconds
Memory usage: 20844 (+10424)
lxml.etree.parse done in 0.016 seconds
Memory usage: 27624 (+17204)
drop_whitespace.parse done in 0.015 seconds
Memory usage: 24468 (+14052)
lxml.etree.XMLParser.feed(): 25317 nodes read in 0.022 seconds
Memory usage: 29844 (+19424)
minidom tree read in 0.288 seconds
Memory usage: 28788 (+18368)
As can be seen from the sizes, both lxml.etree and cElementTree are
rather memory friendly compared to the pure Python libraries
ElementTree and (especially) minidom. Comparing to older CPython
versions, the memory footprint of the minidom library was considerably
reduced in CPython 3.3, by about a factor of 4 in this case.
For plain parser performance, lxml.etree and cElementTree tend to stay
rather close to each other, usually within a factor of two, with
winners well distributed over both sides. Similar timings can be
observed for the ``iterparse()`` function::
lxe: iterparse_bytesIO (SAXR T1) 17.9198 msec/pass
cET: iterparse_bytesIO (SAXR T1) 14.4982 msec/pass
lxe: iterparse_bytesIO (UAXR T3) 8.8522 msec/pass
cET: iterparse_bytesIO (UAXR T3) 12.9857 msec/pass
However, if you benchmark the complete round-trip of a serialise-parse
cycle, the numbers will look similar to these::
lxe: write_utf8_parse_bytesIO (S-TR T1) 19.8867 msec/pass
cET: write_utf8_parse_bytesIO (S-TR T1) 80.7259 msec/pass
lxe: write_utf8_parse_bytesIO (UATR T2) 23.7896 msec/pass
cET: write_utf8_parse_bytesIO (UATR T2) 98.0766 msec/pass
lxe: write_utf8_parse_bytesIO (S-TR T3) 3.0684 msec/pass
cET: write_utf8_parse_bytesIO (S-TR T3) 24.6122 msec/pass
lxe: write_utf8_parse_bytesIO (SATR T4) 0.3495 msec/pass
cET: write_utf8_parse_bytesIO (SATR T4) 1.9610 msec/pass
For applications that require a high parser throughput of large files,
and that do little to no serialization, both cET and lxml.etree are a
good choice. The cET library is particularly fast for iterparse
applications that extract small amounts of data or aggregate
information from large XML data sets that do not fit into memory. If
it comes to round-trip performance, however, lxml is multiple times
faster in total. So, whenever the input documents are not
considerably larger than the output, lxml is the clear winner.
Regarding HTML parsing, Ian Bicking has done some `benchmarking on
lxml's HTML parser`_, comparing it to a number of other famous HTML
parser tools for Python. lxml wins this contest by quite a length.
To give an idea, the numbers suggest that lxml.html can run a couple
of parse-serialise cycles in the time that other tools need for
parsing alone. The comparison even shows some very favourable results
regarding memory consumption.
.. _`benchmarking on lxml's HTML parser`: http://blog.ianbicking.org/2008/03/30/python-html-parser-performance/
Liza Daly has written an article that presents a couple of tweaks to
get the most out of lxml's parser for very large XML documents. She
quite favourably positions ``lxml.etree`` as a tool for
`high-performance XML parsing`_.
.. _`high-performance XML parsing`: http://www.ibm.com/developerworks/xml/library/x-hiperfparse/
Finally, `xml.com`_ has a couple of publications about XML parser
performance. Farwick and Hafner have written two interesting articles
that compare the parser of libxml2 to some major Java based XML
parsers. One deals with `event-driven parser performance`_, the other
one presents `benchmark results comparing DOM parsers`_. Both
comparisons suggest that libxml2's parser performance is largely
superior to all commonly used Java parsers in almost all cases. Note
that the C parser benchmark results are based on xmlbench_, which uses
a simpler setup for libxml2 than lxml does.
.. _`xml.com`: http://www.xml.com/
.. _`event-driven parser performance`: http://www.xml.com/lpt/a/1702
.. _`benchmark results comparing DOM parsers`: http://www.xml.com/lpt/a/1703
.. _xmlbench: http://xmlbench.sourceforge.net/
The ElementTree API
===================
Since all three libraries implement the same API, their performance is
easy to compare in this area. A major disadvantage for lxml's
performance is the different tree model that underlies libxml2. It
allows lxml to provide parent pointers for elements and full XPath
support, but also increases the overhead of tree building and
restructuring. This can be seen from the tree setup times of the
benchmark (given in seconds)::
lxe: -- S- U- -A SA UA
T1: 0.0299 0.0343 0.0344 0.0293 0.0345 0.0342
T2: 0.0368 0.0423 0.0418 0.0427 0.0474 0.0459
T3: 0.0088 0.0084 0.0086 0.0251 0.0258 0.0261
T4: 0.0002 0.0002 0.0002 0.0005 0.0006 0.0006
cET: -- S- U- -A SA UA
T1: 0.0050 0.0045 0.0093 0.0044 0.0043 0.0043
T2: 0.0073 0.0075 0.0074 0.0201 0.0075 0.0074
T3: 0.0033 0.0213 0.0032 0.0034 0.0033 0.0035
T4: 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000
The timings are somewhat close to each other, although cET can be
several times faster than lxml for larger trees. One of the
reasons is that lxml must encode incoming string data and tag names
into UTF-8, and additionally discard the created Python elements
after their use, when they are no longer referenced. ElementTree
represents the tree itself through these objects, which reduces
the overhead in creating them.
Child access
------------
The same tree overhead makes operations like collecting children as in
``list(element)`` more costly in lxml. Where cET can quickly create
a shallow copy of their list of children, lxml has to create a Python
object for each child and collect them in a list::
lxe: root_list_children (--TR T1) 0.0038 msec/pass
cET: root_list_children (--TR T1) 0.0010 msec/pass
lxe: root_list_children (--TR T2) 0.0455 msec/pass
cET: root_list_children (--TR T2) 0.0050 msec/pass
This handicap is also visible when accessing single children::
lxe: first_child (--TR T2) 0.0424 msec/pass
cET: first_child (--TR T2) 0.0384 msec/pass
lxe: last_child (--TR T1) 0.0477 msec/pass
cET: last_child (--TR T1) 0.0467 msec/pass
... unless you also add the time to find a child index in a bigger
list. ET and cET use Python lists here, which are based on arrays.
The data structure used by libxml2 is a linked tree, and thus, a
linked list of children::
lxe: middle_child (--TR T1) 0.0710 msec/pass
cET: middle_child (--TR T1) 0.0420 msec/pass
lxe: middle_child (--TR T2) 1.7393 msec/pass
cET: middle_child (--TR T2) 0.0396 msec/pass
Element creation
----------------
As opposed to ET, libxml2 has a notion of documents that each element must be
in. This results in a major performance difference for creating independent
Elements that end up in independently created documents::
lxe: create_elements (--TC T2) 1.0045 msec/pass
cET: create_elements (--TC T2) 0.0753 msec/pass
Therefore, it is always preferable to create Elements for the document they
are supposed to end up in, either as SubElements of an Element or using the
explicit ``Element.makeelement()`` call::
lxe: makeelement (--TC T2) 1.0586 msec/pass
cET: makeelement (--TC T2) 0.1483 msec/pass
lxe: create_subelements (--TC T2) 0.8826 msec/pass
cET: create_subelements (--TC T2) 0.0827 msec/pass
So, if the main performance bottleneck of an application is creating large XML
trees in memory through calls to Element and SubElement, cET is the best
choice. Note, however, that the serialisation performance may even out this
advantage, especially for smaller trees and trees with many attributes.
Merging different sources
-------------------------
A critical action for lxml is moving elements between document contexts. It
requires lxml to do recursive adaptations throughout the moved tree structure.
The following benchmark appends all root children of the second tree to the
root of the first tree::
lxe: append_from_document (--TR T1,T2) 1.0812 msec/pass
cET: append_from_document (--TR T1,T2) 0.1104 msec/pass
lxe: append_from_document (--TR T3,T4) 0.0155 msec/pass
cET: append_from_document (--TR T3,T4) 0.0060 msec/pass
Although these are fairly small numbers compared to parsing, this easily shows
the different performance classes for lxml and (c)ET. Where the latter do not
have to care about parent pointers and tree structures, lxml has to deep
traverse the appended tree. The performance difference therefore increases
with the size of the tree that is moved.
This difference is not always as visible, but applies to most parts of the
API, like inserting newly created elements::
lxe: insert_from_document (--TR T1,T2) 3.9763 msec/pass
cET: insert_from_document (--TR T1,T2) 0.1459 msec/pass
or replacing the child slice by a newly created element::
lxe: replace_children_element (--TC T1) 0.0749 msec/pass
cET: replace_children_element (--TC T1) 0.0081 msec/pass
as opposed to replacing the slice with an existing element from the
same document::
lxe: replace_children (--TC T1) 0.0052 msec/pass
cET: replace_children (--TC T1) 0.0036 msec/pass
While these numbers are too small to provide a major performance
impact in practice, you should keep this difference in mind when you
merge very large trees. Note that Elements have a ``makeelement()``
method that allows to create an Element within the same document,
thus avoiding the merge overhead when inserting it into that tree.
deepcopy
--------
Deep copying a tree is fast in lxml::
lxe: deepcopy_all (--TR T1) 3.1650 msec/pass
cET: deepcopy_all (--TR T1) 53.9973 msec/pass
lxe: deepcopy_all (-ATR T2) 3.7365 msec/pass
cET: deepcopy_all (-ATR T2) 61.6267 msec/pass
lxe: deepcopy_all (S-TR T3) 0.7913 msec/pass
cET: deepcopy_all (S-TR T3) 13.6220 msec/pass
So, for example, if you have a database-like scenario where you parse in a
large tree and then search and copy independent subtrees from it for further
processing, lxml is by far the best choice here.
Tree traversal
--------------
Another important area in XML processing is iteration for tree
traversal. If your algorithms can benefit from step-by-step
traversal of the XML tree and especially if few elements are of
interest or the target element tag name is known, the ``.iter()``
method is a good choice::
lxe: iter_all (--TR T1) 1.0529 msec/pass
cET: iter_all (--TR T1) 0.2635 msec/pass
lxe: iter_islice (--TR T2) 0.0110 msec/pass
cET: iter_islice (--TR T2) 0.0050 msec/pass
lxe: iter_tag (--TR T2) 0.0079 msec/pass
cET: iter_tag (--TR T2) 0.0112 msec/pass
lxe: iter_tag_all (--TR T2) 0.1822 msec/pass
cET: iter_tag_all (--TR T2) 0.5343 msec/pass
This translates directly into similar timings for ``Element.findall()``::
lxe: findall (--TR T2) 1.7176 msec/pass
cET: findall (--TR T2) 0.9973 msec/pass
lxe: findall (--TR T3) 0.3967 msec/pass
cET: findall (--TR T3) 0.2525 msec/pass
lxe: findall_tag (--TR T2) 0.2258 msec/pass
cET: findall_tag (--TR T2) 0.5770 msec/pass
lxe: findall_tag (--TR T3) 0.1085 msec/pass
cET: findall_tag (--TR T3) 0.1919 msec/pass
Note that all three libraries currently use the same Python
implementation for ``.findall()``, except for their native tree
iterator (``element.iter()``). In general, lxml is very fast
for iteration, but loses ground against cET when many Elements
are found and need to be instantiated. So, the more selective
your search is, the faster lxml will run.
XPath
=====
The following timings are based on the benchmark script `bench_xpath.py`_.
This part of lxml does not have an equivalent in ElementTree. However, lxml
provides more than one way of accessing it and you should take care which part
of the lxml API you use. The most straight forward way is to call the
``xpath()`` method on an Element or ElementTree::
lxe: xpath_method (--TC T1) 0.3982 msec/pass
lxe: xpath_method (--TC T2) 7.8895 msec/pass
lxe: xpath_method (--TC T3) 0.0477 msec/pass
lxe: xpath_method (--TC T4) 0.3982 msec/pass
This is well suited for testing and when the XPath expressions are as diverse
as the trees they are called on. However, if you have a single XPath
expression that you want to apply to a larger number of different elements,
the ``XPath`` class is the most efficient way to do it::
lxe: xpath_class (--TC T1) 0.0713 msec/pass
lxe: xpath_class (--TC T2) 1.1325 msec/pass
lxe: xpath_class (--TC T3) 0.0215 msec/pass
lxe: xpath_class (--TC T4) 0.0722 msec/pass
Note that this still allows you to use variables in the expression, so you can
parse it once and then adapt it through variables at call time. In other
cases, where you have a fixed Element or ElementTree and want to run different
expressions on it, you should consider the ``XPathEvaluator``::
lxe: xpath_element (--TR T1) 0.1101 msec/pass
lxe: xpath_element (--TR T2) 2.0473 msec/pass
lxe: xpath_element (--TR T3) 0.0267 msec/pass
lxe: xpath_element (--TR T4) 0.1087 msec/pass
While it looks slightly slower, creating an XPath object for each of the
expressions generates a much higher overhead here::
lxe: xpath_class_repeat (--TC T1 ) 0.3884 msec/pass
lxe: xpath_class_repeat (--TC T2 ) 7.6182 msec/pass
lxe: xpath_class_repeat (--TC T3 ) 0.0465 msec/pass
lxe: xpath_class_repeat (--TC T4 ) 0.3877 msec/pass
Note that tree iteration can be substantially faster than XPath if
your code short-circuits after the first couple of elements were
found. The XPath engine will always return the complete result set,
regardless of how much of it will actually be used.
Here is an example where only the first matching element is being
searched, a case for which XPath has syntax support as well::
lxe: find_single (--TR T2) 0.0184 msec/pass
cET: find_single (--TR T2) 0.0052 msec/pass
lxe: iter_single (--TR T2) 0.0024 msec/pass
cET: iter_single (--TR T2) 0.0007 msec/pass
lxe: xpath_single (--TR T2) 0.0033 msec/pass
When looking for the first two elements out of many, the numbers
explode for XPath, as restricting the result subset requires a
more complex expression::
lxe: iterfind_two (--TR T2) 0.0184 msec/pass
cET: iterfind_two (--TR T2) 0.0062 msec/pass
lxe: iter_two (--TR T2) 0.0029 msec/pass
cET: iter_two (--TR T2) 0.0017 msec/pass
lxe: xpath_two (--TR T2) 0.2768 msec/pass
A longer example
================
... based on lxml 1.3.
A while ago, Uche Ogbuji posted a `benchmark proposal`_ that would
read in a 3MB XML version of the `Old Testament`_ of the Bible and
look for the word *begat* in all verses. Apparently, it is contained
in 120 out of almost 24000 verses. This is easy to implement in
ElementTree using ``findall()``. However, the fastest and most memory
friendly way to do this is obviously ``iterparse()``, as most of the
data is not of any interest.
.. _`benchmark proposal`: http://www.onlamp.com/pub/wlg/6291
.. _`Old Testament`: http://www.ibiblio.org/bosak/xml/eg/religion.2.00.xml.zip
Now, Uche's original proposal was more or less the following:
.. sourcecode:: python
def bench_ET():
tree = ElementTree.parse("ot.xml")
result = []
for v in tree.findall("//v"):
text = v.text
if 'begat' in text:
result.append(text)
return len(result)
which takes about one second on my machine today. The faster ``iterparse()``
variant looks like this:
.. sourcecode:: python
def bench_ET_iterparse():
result = []
for event, v in ElementTree.iterparse("ot.xml"):
if v.tag == 'v':
text = v.text
if 'begat' in text:
result.append(text)
v.clear()
return len(result)
The improvement is about 10%. At the time I first tried (early 2006), lxml
didn't have ``iterparse()`` support, but the ``findall()`` variant was already
faster than ElementTree. This changes immediately when you switch to
cElementTree. The latter only needs 0.17 seconds to do the trick today and
only some impressive 0.10 seconds when running the iterparse version. And
even back then, it was quite a bit faster than what lxml could achieve.
Since then, lxml has matured a lot and has gotten much faster. The iterparse
variant now runs in 0.14 seconds, and if you remove the ``v.clear()``, it is
even a little faster (which isn't the case for cElementTree).
One of the many great tools in lxml is XPath, a Swiss army knife for finding
things in XML documents. It is possible to move the whole thing to a pure
XPath implementation, which looks like this:
.. sourcecode:: python
def bench_lxml_xpath_all():
tree = etree.parse("ot.xml")
result = tree.xpath("//v[contains(., 'begat')]/text()")
return len(result)
This runs in about 0.13 seconds and is about the shortest possible
implementation (in lines of Python code) that I could come up with. Now, this
is already a rather complex XPath expression compared to the simple "//v"
ElementPath expression we started with. Since this is also valid XPath, let's
try this instead:
.. sourcecode:: python
def bench_lxml_xpath():
tree = etree.parse("ot.xml")
result = []
for v in tree.xpath("//v"):
text = v.text
if 'begat' in text:
result.append(text)
return len(result)
This gets us down to 0.12 seconds, thus showing that a generic XPath
evaluation engine cannot always compete with a simpler, tailored solution.
However, since this is not much different from the original findall variant,
we can remove the complexity of the XPath call completely and just go with
what we had in the beginning. Under lxml, this runs in the same 0.12 seconds.
But there is one thing left to try. We can replace the simple ElementPath
expression with a native tree iterator:
.. sourcecode:: python
def bench_lxml_getiterator():
tree = etree.parse("ot.xml")
result = []
for v in tree.getiterator("v"):
text = v.text
if 'begat' in text:
result.append(text)
return len(result)
This implements the same thing, just without the overhead of parsing and
evaluating a path expression. And this makes it another bit faster, down to
0.11 seconds. For comparison, cElementTree runs this version in 0.17 seconds.
So, what have we learned?
* Python code is not slow. The pure XPath solution was not even as fast as
the first shot Python implementation. In general, a few more lines in
Python make things more readable, which is much more important than the last
5% of performance.
* It's important to know the available options - and it's worth starting with
the most simple one. In this case, a programmer would then probably have
started with ``getiterator("v")`` or ``iterparse()``. Either of them would
already have been the most efficient, depending on which library is used.
* It's important to know your tool. lxml and cElementTree are both very fast
libraries, but they do not have the same performance characteristics. The
fastest solution in one library can be comparatively slow in the other. If
you optimise, optimise for the specific target platform.
* It's not always worth optimising. After all that hassle we got from 0.12
seconds for the initial implementation to 0.11 seconds. Switching over to
cElementTree and writing an ``iterparse()`` based version would have given
us 0.10 seconds - not a big difference for 3MB of XML.
* Take care what operation is really dominating in your use case. If we split
up the operations, we can see that lxml is slightly slower than cElementTree
on ``parse()`` (both about 0.06 seconds), but more visibly slower on
``iterparse()``: 0.07 versus 0.10 seconds. However, tree iteration in lxml
is incredibly fast, so it can be better to parse the whole tree and then
iterate over it rather than using ``iterparse()`` to do both in one step.
Or, you can just wait for the lxml developers to optimise iterparse in one
of the next releases...
lxml.objectify
==============
The following timings are based on the benchmark script `bench_objectify.py`_.
Objectify is a data-binding API for XML based on lxml.etree, that was added in
version 1.1. It uses standard Python attribute access to traverse the XML
tree. It also features ObjectPath, a fast path language based on the same
meme.
Just like lxml.etree, lxml.objectify creates Python representations of
elements on the fly. To save memory, the normal Python garbage collection
mechanisms will discard them when their last reference is gone. In cases
where deeply nested elements are frequently accessed through the objectify
API, the create-discard cycles can become a bottleneck, as elements have to be
instantiated over and over again.
ObjectPath
----------
ObjectPath can be used to speed up the access to elements that are deep in the
tree. It avoids step-by-step Python element instantiations along the path,
which can substantially improve the access time::
lxe: attribute (--TR T1) 4.1828 msec/pass
lxe: attribute (--TR T2) 17.3802 msec/pass
lxe: attribute (--TR T4) 3.8657 msec/pass
lxe: objectpath (--TR T1) 0.9289 msec/pass
lxe: objectpath (--TR T2) 13.3109 msec/pass
lxe: objectpath (--TR T4) 0.9289 msec/pass
lxe: attributes_deep (--TR T1) 6.2900 msec/pass
lxe: attributes_deep (--TR T2) 20.4713 msec/pass
lxe: attributes_deep (--TR T4) 6.1679 msec/pass
lxe: objectpath_deep (--TR T1) 1.3049 msec/pass
lxe: objectpath_deep (--TR T2) 14.0815 msec/pass
lxe: objectpath_deep (--TR T4) 1.3051 msec/pass
Note, however, that parsing ObjectPath expressions is not for free either, so
this is most effective for frequently accessing the same element.
Caching Elements
----------------
A way to improve the normal attribute access time is static instantiation of
the Python objects, thus trading memory for speed. Just create a cache
dictionary and run:
.. sourcecode:: python
cache[root] = list(root.iter())
after parsing and:
.. sourcecode:: python
del cache[root]
when you are done with the tree. This will keep the Python element
representations of all elements alive and thus avoid the overhead of repeated
Python object creation. You can also consider using filters or generator
expressions to be more selective. By choosing the right trees (or even
subtrees and elements) to cache, you can trade memory usage against access
speed::
lxe: attribute_cached (--TR T1) 3.1357 msec/pass
lxe: attribute_cached (--TR T2) 15.8911 msec/pass
lxe: attribute_cached (--TR T4) 2.9194 msec/pass
lxe: attributes_deep_cached (--TR T1) 3.8984 msec/pass
lxe: attributes_deep_cached (--TR T2) 16.8300 msec/pass
lxe: attributes_deep_cached (--TR T4) 3.6936 msec/pass
lxe: objectpath_deep_cached (--TR T1) 0.7496 msec/pass
lxe: objectpath_deep_cached (--TR T2) 12.3763 msec/pass
lxe: objectpath_deep_cached (--TR T4) 0.7427 msec/pass
Things to note: you cannot currently use ``weakref.WeakKeyDictionary`` objects
for this as lxml's element objects do not support weak references (which are
costly in terms of memory). Also note that new element objects that you add
to these trees will not turn up in the cache automatically and will therefore
still be garbage collected when all their Python references are gone, so this
is most effective for largely immutable trees. You should consider using a
set instead of a list in this case and add new elements by hand.
Further optimisations
---------------------
Here are some more things to try if optimisation is required:
* A lot of time is usually spent in tree traversal to find the addressed
elements in the tree. If you often work in subtrees, do what you would also
do with deep Python objects: assign the parent of the subtree to a variable
or pass it into functions instead of starting at the root. This allows
accessing its descendants more directly.
* Try assigning data values directly to attributes instead of passing them
through DataElement.
* If you use custom data types that are costly to parse, try running
``objectify.annotate()`` over read-only trees to speed up the attribute type
inference on read access.
Note that none of these measures is guaranteed to speed up your application.
As usual, you should prefer readable code over premature optimisations and
profile your expected use cases before bothering to apply optimisations at
random.