Project

Profile

Help

Forums » Help »

Performance of DOM Wrapper classes

Added by Anonymous almost 17 years ago

Legacy ID: #4463120 Legacy Poster: Colin Green (colgreen)

Hi Michael, In your blog post "Saxon Vs Xalan Performance; and DOM" dated 31/3/07 (and subsequent updates) you mention that you modified the Java DOM wrapper classes so that they no longer perform an eager iteration over an axis - placing wrapped result nodes into an array. However I noticed that the latest version of these wrapper classes (both the Java and .Net DOM wrappers) are in fact still operating in the old, slow way - allocating an array and placing wrapped nodes into it. I checked for updates on SVN but no luck. Soooo, I was wondering whether these updates are perhaps Saxon-SA only or if they just haven't been submitted yet? Or am I just looking in the wrong place! Thanks in advance, Colin Green, Focus Business Solutions Ltd.


Replies (2)

RE: Performance of DOM Wrapper classes - Added by Anonymous almost 17 years ago

Legacy ID: #4464521 Legacy Poster: Michael Kay (mhkay)

I stick to a very strict policy of only fixing bugs in maintenance releases - no performance or other improvements, because of the risk of introducing new bugs. Unfortunately it's been a long while since the last functionality release, and these changes are queued up with many others waiting until the next one happens. (And sorry, no dates: I make a policy of not announcing dates. In fact, I don't even schedule dates; I find this is a good discipline that means I don't fall into the temptation of releasing code that isn't actually reliable enough yet.) Michael Kay

RE: Performance of DOM Wrapper classes - Added by Anonymous almost 17 years ago

Legacy ID: #4464548 Legacy Poster: Colin Green (colgreen)

Hi, Thanks for the info. Yes I fully appreciate that deadlines/schedules can result in rushed and therefore buggy code. In actual fact this isn't a problem for us /right/ now and knowing that the change will filter down into a future release is really all we need to know. If it did become a problem we could probably make similar modifications here as a temporary measure - looking at the code it looks like a fairly straightforward exercise. Thanks again and thanks for a great product. Colin Green

    (1-2/2)

    Please register to reply