Lists Home |
Date Index |
I think you are absolutely correct, the tag to text ratio definitely
has an impact. I wrote a very simplistic dom implementation (in C,
loaded by expat) and tested it
with a machine generated file with a lot of empty tags. This had a big
to a file that is mostly text. Haven't tried that in Java, but it might
Liam Quin wrote:
>On Tue, Apr 06, 2004 at 08:01:42AM -0400, Mark Schmeets wrote:
>>I have not tried xmlbuddy, but all of the xml editors I have tried,
>>including oXygen ( and the Win32 ones XMLSpy, XMLNotepad, etc. ) exhibit
>>this behavior. Using RAM at 10 to 12 times the size of the document on
>>disk. I have always assumed it was because the were using DOM's
>>internally, and that the
>>DOM implementations were memory hogs. Any one have any wisdom to share here?
>Have you tested how the density of tags to text affects this?
>I remember that with SoftQuad Author/Editor (in SGML days), the
>overhead per element was significant, but text averaged not so
>much more than disk overhead. Systems using 16-bit characters
>internally to represent Unicode text will do slightly worse than
>using twice as much memory as disk space for text that's mostly
>For Java-based software, the JVM you use may also make a difference,
>although my experience (mostly on Linux) has been that the non-Sun JVMs
>don't actually run much software I want to run.