php|architect's Guide to Web Scraping with PHP - Wind Business ...

100 ” DOM Extension
T y p of P e a r s e r s
Before going much further, you should be aware that there are two types of XML
parsers: tree parsers and pull parsers. T r e parsers e load the entire document into
memory and allow you to access any part of it at any time as well as manipulate it.
Pull parsers read the document a piece at a time and limit you to working with the
current piece being read.
The two types of parsers share a relationship similar to that between the
file_get_contents and fgets functions: the former lets you work with the entire document
at once and uses as much memory needed to store it, while the latter allows
you to work with a piece of the document at a time and use less memory in the process.
When working with fairly large documents, lower memory usage is generally the
preferable option. Attempting to load a huge document into memory all at once has
the same effect on the local system as a throttling client does on a web server: in
both cases, resources are consumed and system performance is debilitated until the
system eventually locks up or crashes under the stress.
The DOM extension is a tree parser. In general, web scraping does not require
the ability to access all parts of the document simultaneously. H o w e v e r, the type of
data extraction involved in web scraping can be rather extensive to implement using
a pull parser. The appropriateness of extension o v e r the other depends on the size
and complexity of the document.
Loading Documents
The DOMDocument class is where use of the DOM extension begins. The first thing to
do is instantiate it and then feed it the validated markup data. N o t e that the DOM extension
will emit warnings when a document is loaded if that document is not valid
or well-formed. T oavoid this, see the previous chapter on using the tidy extension.
If tidy does not eliminate the issue, errors can be controlled as shown in the example
below. N o t e that errors are buffered until manually cleared, so make a point of clearing
them after each load operation if they are not needed to avoid wasting memory.