[Exist-open] XUpdate news

[Wolfgang M. <me...@if...>]

Hi,

I'm happy to announce that the CVS version does now implement most of the=
=20
XUpdate spec. All commands are supported (append, remove, rename, update,=
=20
insert-after, insert-before).=20

Indexes are now correctly maintained. A partial document re-index is=20
automatically triggered if required. Basically, there are two different k=
inds=20
of re-index runs:

1) update, insert and remove operations trigger a re-index of the current=
 node=20
and its descendant nodes.

2) furthermore, if a append, update or insert operation changes the overa=
ll=20
structure of the document tree (i.e. it changes the order of the tree at =
some=20
level), all levels of the document tree located below the level where the=
=20
modification occurred have to be reindexed.

This sounds complicated, but it seems to work even better than I had expe=
cted.=20
In particular, the second type of re-index run should not occur too often=
=2E=20
Since many documents tend to have a rather irregular structure, there wil=
l=20
usually be some spare identifiers between nodes, which can be used for th=
e=20
update. A document re-index is thus not needed.

Todo: variables work only if declared at the start of the xupdate doc (I =
have=20
no idea why), comments and processing-instructions are not correctly hand=
led.=20
I also plan to add a configuration option to insert additional spare ids=20
between nodes when storing the document, thus reducing the frequency of=20
document re-index runs.

With the CVS version, you can try xupdate using the client: first, insert=
 the=20
document samples/xupdate/address.xml into collection /db/test. Then do

bin/client.sh -l -c /db/test -X samples/xupdate/xupdate.xml

which will apply a number of different modifications to the document. Ple=
ase=20
note that the xupdate is applied to the entire collection specified with =
-c.

Best regards,

Wolfgang