Element.getElementsByTagName Efficiency 2003-01-22 - By Joseph Kesselman
> what is the underlying architecture of the Element class and its impact
> on getElementsByTagName
Xerces is open source, and the DOM implementation isn't particularly
complicated code; reading the code might be the best way to get the
details you're looking for.
(I could discuss how it used to work -- basic tree-walk looking for
matching nodes, plus caching for faster re-access, plus an extremely
annoying cache flush to match the DOM's requirement that Nodelist be a
"live view" -- but I haven't checked the implementation to see if it still
works that way. I'd expect it does, though.)
You'll probably also want to look at the DOM Level 2 TreeWalker and
NodeIterator as possible alternatives. With appropriate filtering they can
perform the same kind of search, more efficiently if your filter
understands how to skip irrelevant subtrees, and they avoid most or all of
the cache/flush overhead.
______________________________________
Joe Kesselman / IBM Research
<br><font size=2 face="sans-serif">></font><font size=2><tt> what is
the underlying architecture of the Element class and its impact</tt></font>
<br><font size=2><tt>> on getElementsByTagName</tt></font>
<br>
<br><font size=2><tt>Xerces is open source, and the DOM implementation
isn't particularly complicated code; reading the code might be the best
way to get the details you're looking for. </tt></font>
<br>
<br><font size=2><tt>(I could discuss how it used to work -- basic tree-walk
looking for matching nodes, plus caching for faster re-access, plus an
extremely annoying cache flush to match the DOM's requirement that Nodelist
be a "live view" -- but I haven't checked the implementation
to see if it still works that way. I'd expect it does, though.)</tt></font>
<br>
<br><font size=2><tt>You'll probably also want to look at the DOM Level
2 TreeWalker and NodeIterator as possible alternatives. With appropriate
filtering they can perform the same kind of search, more efficiently if
your filter understands how to skip irrelevant subtrees, and they avoid
most or all of the cache/flush overhead.</tt></font>
<br><font size=2 face="sans-serif"><br>
______________________________________<br>
Joe Kesselman / IBM Research<br>
</font>
|
|
