On 5 Jul 2001, at 10:27, Carsten Ziegeler wrote:
> Hi,
>
> with the current caching algorithm implemented in Cocoon2 it is
> not possible to cache either the CInludeTransformer nor the
> XIncludeTransformer.
>
It is possible, although hacky.
> Why is this so?
>
> Before the sax stream is generated (before the generator starts
> reading its xml document), the caching algorithm builds the
> key and collects the validity objects for the current request.
>
> At this stage, the CachingCIncludeTransformer didn't get any
> startElement() call, so it does not know which content will
> be included at a later time. So the key and the validity
> objects cannot contain any information about this.
>
CachingCIncludeTransformer generates "empty" IncludeCacheValidity:
<code>
public CacheValidity generateValidity() {
try {
currentCacheValidity = new IncludeCacheValidity(sourceResolver);
return currentCacheValidity;
} catch (Exception e) {
getLogger().error("CachingCIncludeTransformer: could not generateKey", e);
return null;
}
}
</code>
and fills it with data during transformation:
<code>
protected void processCIncludeElement(
String src, String element, String ns, String prefix)
{
try {
currentCacheValidity.add(src,
sourceResolver.resolve(src).getLastModified());
...
</code>
Validity of IncludeCacheValidity is calculated not by comparing if with new
IncludeCacheValidity object (as in AggregatedCacheValidity) but by comparing
timestamps:
<code>
public boolean isValid(CacheValidity validity) {
if (validity instanceof IncludeCacheValidity) {
SourceResolver otherResolver = ((IncludeCacheValidity) validity).resolver;
for(Iterator i = sources.iterator(), j = timeStamps.iterator();
i.hasNext();) {
String src = ((String)i.next());
long timeStamp = ((Long)j.next()).longValue();
try {
if( otherResolver.resolve(src).getLastModified()!= timeStamp )
return false;
} catch (Exception e) {
return false;
}
}
return true;
}
return false;
}
</code>
CachingCIncludeTransformer always generates the same key since which documents
are included depends only on former generation/transformation stages:
<code>
public long generateKey() {
return 1;
}
</code>
> So the key and your validity objects your transformer generates
> are always the same, regardless which documents are included.
>
> Sorry.
Not at all ;->
>
> Regarding caching in content aggregation:
> It is implemented and should work, if you do not aggregate pipelines. If you
> aggregate xml files it is cached. In opposite to the Transformer the content
> aggregation knows beforehand which documents are aggregated (or included if you
> like), so the key and the validity objects generated contain all this
> information.
>
I am eager to work on aggregation of pipelines. It seems that some redesign in
SourceResolver and Source concepts is necessary. Any suggestions?
Maciek Kaminski
[EMAIL PROTECTED]
---------------------------------------------------------------------
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, email: [EMAIL PROTECTED]
