Re: [Fwd: XML_Indexing proposal]

Hi Bertrand, bertrand Gugger wrote:

XML_Indexing does /not exactly supports XPath/. It uses an XPath like syntax, to easily locate a subset of the whole XML document. The rule, in this regard, is that any of the supported expressions must be valid XPath, but the opposit isn't true.

< You could say: "supports the most usefull subset of XPath's syntax"

I'd say : "supports the most usefull subset of XPath's syntax for what a big XML file is expected to contain (ie: a lot of repetitive blocks, as RDF)"

< usefull=80% < what 80% you mean ?

"The idea is to address about 80% of the needs when it comes to accessing big xml files." I expect (around) 80% of _big_ xml files to require simple XPath statements.

Examples are there but not what the input is ('test.xml') nor what output should.

Actually I used (mozilla's) RDF files for testing and benchmarking, as mentioned in the Benchmark section. The output, for now, is raw extracted XML data, but I may provide support for direct extraction of DOM objects.

OK for bench: what machine, what disk, what is millisecond ?

The benchmark shows "speed ratio", for example with a 2 MB file, XML_Indexing+DOM is 40 times faster than pure DOM. It is not absolute timings, which would be dependent on my system disk/cpu/etc... but relative "ratios". However, for information my test system is : - Linux Debian Woody / PHP 4.3.4 / PHP 5.0.2 - Single-CPU Duron 700Mhz - 320 Mb RAM - IDE disk for which hdparm reports : 31 MB/s

In the code, what is the need for circle include/require the base XML/indexing ?

I mean         include_once 'XML/Indexing/Builder/Numeric.php';
in         'XML/Indexing/Builder.php';   and then

require_once 'XML/Indexing/Builder.php';

in         'XML/Indexing/Builder/Numeric.php';

The call to include_once 'XML/Indexing/Builder/Numeric.php' in XML/Indexing/Builder.php is part of a workaround for a bug in PHP5 expat lib, and therefore is only performed with PHP 5.x versions. For normal operation, one should always instantiate the driver class directly, that is : XML_Indexing_Builder_Numeric. Thanks for your review :-) But, the proposal and voting periods did end, and my post concerns the process of establishing XML_Indexing as an official PEAR Package. -- og