Replace libxml++ w/ libxml2 #218

rakhimov · 2017-08-21T20:19:16Z

XML parser facilities are only needed to process MEF and config files.
SCRAM does not need all the bells & whistles provided by libxml++.
It should be possible to replace libxml++
with a custom minimal adapter around libxml2,
providing only the needed functionality.
This would free SCRAM from libxml++ and its dependencies
(glib, glibmm, libsigc++).

Moreover, the performance should improve slightly
if the custom wrapper does not manage wrapper node databases
and XML modification as libxml++ does.

The codebase needs to be updated for the new libxml++3 API anyway,
so this is a good excuse for creating the custom adaptor and dropping libxml++.

Required features:

DOM parser & read-only document (from file & stream)
XInclude processing (automatic w/ libxml2)
RelaxNG validator
Element nodes only (w/ attribute & text values)

The text was updated successfully, but these errors were encountered:

This class wraps libxml++ API for convenience. The minimal API serves only the needs of the SCRAM code, which doesn't need many functionalities provided by libxml++. This and other xml helper classes should hide the complexity of XML libraries. Issue #218

The libxml2 XML document tracks its filename, so it can be used instead of tracking files separately in a table. Issue #218

The standard XML Node API can be used when the node type is known to be Element. This relies on validated XML input. Unfortunately, since libxml++ uses std::list<Node*> for children, the performance degrades by around 5% compared to XPath queries returning std::vector<Node*>. Issue #218

XML child elements are intrusive linked lists, so it is convenient to reuse this data structure instead of converting it into C++ std::list<>. The range interface just wraps the begin and end nodes, as well as ensuring that child nodes are actually Element nodes. Issue #218

xml::Element fully replaces xmlpp::Element*. Config does not need as much use as Initializer, so this is an initial indirect test of the xml::Element API. Issue #218

libxml++ xmlpp::Element, xmlpp::NodeSet, and xmlpp::Node::NodeList are replaced with xml::Element and xml::Element::Range. Initializer private interfaces and code are changed appropriately to accommodate the xml::Element. xml::Element is also refactored to support the functionalities required by Initializer. Issue #218

These are all thin and convenient wrappers for libxml++ classes. Issue #218

Initializer now completely free from libxml++ code. Issue #218

Config is now completely free from libxml++ code. Issue #218

This is a convenience overload to work with C++ input streams. The overload is needed for tests. Issue #218

Tests are now free from libxml++ code. Issue #218

The GUI mainwindow code now is free from libxml++. Issue #218

xml::Parser provides all the needed functionality. Issue #218

xml::Element adaptor directly uses libxml2 facilities. The downside of this approach is that libxml2 C code pollutes the global namespace with C xml functions. Issue #218

Validator, Parser, and Document are now fully implemented in terms of libxml2 code. These are the last facilities to transition to libxml2. Even though the code switched to libxml2, the proper error reporting is not yet implemented. Issue #218

Only libxml2 version is needed. Issue #218

Use the standard FindLibXml2 to find libxml2. Issue #218

Installation of libxml++ is removed. Issue #218

- Remove libxml++ installation instructions. - The minimum version (2.9.1) for libxml2 is chosen to be what is in Ubuntu 14.04. Older versions may also work just fine. This is rather untested. Issue #218

This provides a thin view to the underlying UTF-8 strings without excessive copying them to std::string. Issue #218

Issue #218

This is way faster than boost::lexical_cast. The overall MEF initialization improves by 10%. Issue #218

This is the library time that needs to be subtracted from overall initialization time. Issue #218

Upon moving to libxml2 and refactoring, the overall gain is 2 - 2.5x speedup. However, there's only 1-2% memory utilization improvement w/ libxml2. Issue #218

GUI does not need to parse the files again to get the DOM document for validation purposes. Issue #218

rakhimov · 2017-08-23T23:56:05Z

libxml++ is fully replaced by libxml2 custom wrapper,
but error message construction is not handled.
This is delegated to #219 .

rakhimov added in progress optimization quality labels Aug 21, 2017

rakhimov added this to the v0.16.0 milestone Aug 21, 2017

rakhimov mentioned this issue Aug 23, 2017

Redesign Exception classes w/ boost::exception #219

Closed

rakhimov added a commit that referenced this issue Aug 23, 2017

Replace Initializer doc_to_file with XML doc->URL

8de2fe7

The libxml2 XML document tracks its filename, so it can be used instead of tracking files separately in a table. Issue #218

rakhimov added a commit that referenced this issue Aug 23, 2017

Refactor Config with xml::Element

74f895f

xml::Element fully replaces xmlpp::Element*. Config does not need as much use as Initializer, so this is an initial indirect test of the xml::Element API. Issue #218

rakhimov added a commit that referenced this issue Aug 23, 2017

Add xml::{Document, Parser, Validator} classes

4e40276

These are all thin and convenient wrappers for libxml++ classes. Issue #218

rakhimov added a commit that referenced this issue Aug 23, 2017

Refactor Initializer w/ xml::{Parser, Validator}

5f1d8bc

Initializer now completely free from libxml++ code. Issue #218

rakhimov added a commit that referenced this issue Aug 23, 2017

Refactor Config w/ xml::{Parser, Validator}

d674271

Config is now completely free from libxml++ code. Issue #218

rakhimov added a commit that referenced this issue Aug 23, 2017

Provide overload xml::Parser to parse input stream

0ab1fbd

This is a convenience overload to work with C++ input streams. The overload is needed for tests. Issue #218

rakhimov added a commit that referenced this issue Aug 23, 2017

Refactor tests with xml::{Parser, Validator}

d61c995

Tests are now free from libxml++ code. Issue #218

rakhimov added a commit that referenced this issue Aug 23, 2017

Refactor GUI code w/ xml::{Parser, Validator}

b0bca4b

The GUI mainwindow code now is free from libxml++. Issue #218

rakhimov added a commit that referenced this issue Aug 23, 2017

Remove ConstructDomParser() in xml.h

5a62895

xml::Parser provides all the needed functionality. Issue #218

rakhimov added a commit that referenced this issue Aug 23, 2017

Replace libxml++ w/ libxml2 in xml::Element impl

cc14155

xml::Element adaptor directly uses libxml2 facilities. The downside of this approach is that libxml2 C code pollutes the global namespace with C xml functions. Issue #218

rakhimov added a commit that referenced this issue Aug 23, 2017

Remove libxml++ version info in code

d7b5de4

Only libxml2 version is needed. Issue #218

rakhimov added a commit that referenced this issue Aug 23, 2017

Remove libxml++ from CMake configurations

6b1372f

Use the standard FindLibXml2 to find libxml2. Issue #218

rakhimov added a commit that referenced this issue Aug 23, 2017

Update CI scripts with libxml2

5a570da

Installation of libxml++ is removed. Issue #218

rakhimov added a commit that referenced this issue Aug 23, 2017

Refactor xml.h facilities w/ string view

cf6afc8

This provides a thin view to the underlying UTF-8 strings without excessive copying them to std::string. Issue #218

rakhimov added a commit that referenced this issue Aug 23, 2017

Move internal XML helpers into xml::detail ns

c184fb6

Issue #218

rakhimov added a commit that referenced this issue Aug 23, 2017

Optimize xml::CastValue w/ cstdlib strtod & strtol

34f9a35

This is way faster than boost::lexical_cast. The overall MEF initialization improves by 10%. Issue #218

rakhimov added a commit that referenced this issue Aug 23, 2017

Add logging of XML parse per file in Initializer

ed571ee

This is the library time that needs to be subtracted from overall initialization time. Issue #218

rakhimov added a commit that referenced this issue Aug 23, 2017

Update Initializer performance records

32ae944

Upon moving to libxml2 and refactoring, the overall gain is 2 - 2.5x speedup. However, there's only 1-2% memory utilization improvement w/ libxml2. Issue #218

rakhimov added a commit that referenced this issue Aug 23, 2017

Provide XML DOM docs to GUI from Initializer

e4bfc33

GUI does not need to parse the files again to get the DOM document for validation purposes. Issue #218

rakhimov closed this as completed Aug 23, 2017

rakhimov removed the in progress label Aug 23, 2017

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Replace libxml++ w/ libxml2 #218

Replace libxml++ w/ libxml2 #218

rakhimov commented Aug 21, 2017 •

edited

Loading

rakhimov commented Aug 23, 2017

Replace libxml++ w/ libxml2 #218

Replace libxml++ w/ libxml2 #218

Comments

rakhimov commented Aug 21, 2017 • edited Loading

rakhimov commented Aug 23, 2017

rakhimov commented Aug 21, 2017 •

edited

Loading