You are viewing a plain text version of this content. The canonical link for it is here.
Posted to oak-dev@jackrabbit.apache.org by Jörg Hoh <jh...@googlemail.com.INVALID> on 2022/01/14 10:03:00 UTC
Check if a node has children NOT matching a pattern
Hi,
I have a requirement to check if a node has child nodes, which do not match
a certain pattern. This is definitely important, because
node.hasChildren()
will return true, if there is a "rep:policy" childnode below it. But this
node is an implementation detail of Oak, and it does not indicate that
there is a "real" child node below this node. The same happens with
"jcr:content" child nodes or other implementation specific nodes, which are
not considered as "independent child nodes" from an application point of
view.
The existing API allows only to check for the presence of child nodes
matching a certain glob (but not "not matching"), so it is of no help for
me.
Right now it is implemented by iterating over the child nodes and returning
when the first child is detected which does not match the above criteria.
This works fine, but it is slow when the child node list is very long (e.g.
hundreds of nodes). Because then just getting the iterator can take
miliseconds, especially when the Oak repo is backed by Mongo.
On top i need to execute this operation quite a number of times within a
single request, so that getting these iterators can consume quite some
time, making my requests slower than they should be.
I would like to have an API similar to the existing API, maybe looking like
this:
node.hasChildren(Pattern notmatchingNodeNames)
which allows me to provide a pattern to it. As soon as a childnode's name
is found not matching the pattern, the function returns true. This would
allow the implementation to iterate only over the minimal number of child
nodes, and I would expect it to be much faster than my existing
implementation for these larger folders.
I know that the JCR API can hardly be extended by this, but having this in
a Util class would be fine as well.
WDYT?
Jörg
--
Cheers,
Jörg Hoh,
https://cqdump.joerghoh.de
Twitter: @joerghoh
Re: Check if a node has children NOT matching a pattern
Posted by Jörg Hoh <jh...@googlemail.com.INVALID>.
Hi Angela,
I raised https://issues.apache.org/jira/browse/JCR-4759 and incorporated
your feedback. Not sure if I can provide an implementation, not familiar
enough with JR/Oak internals :-)
Jörg
Am Fr., 14. Jan. 2022 um 11:45 Uhr schrieb Angela Schreiber
<an...@adobe.com.invalid>:
> Hi Jörg
>
> Thanks a lot for reaching out.
>
> IMHO it makes sense to have this reported as feature request for
> jackrabbit-api as there is no easy way to extend the existing JCR Node API
> contract.
>
> There already exists a JackrabbitNode interface in oak-jackrabbit-api,
> which we could use to add the new functionality.
>
> Would you be able to file a feature request in JIRA for this?
> Obviously, a patch or PR would be welcome as well.
>
> Kind regards
> Angela
>
> PS: i am wondering why the Pattern param would be named
> 'notmatchingNodeNames'. it could be any kind of regex Pattern, couldn't it?
>
> PS2: for consistency we might also consider adding
> JackrabbitNode.getChildren(Pattern pattern) alongside the new hasChildren
> method.
>
>
> ________________________________
> From: Jörg Hoh <jh...@googlemail.com.INVALID>
> Sent: Friday, January 14, 2022 11:03 AM
> To: oak-dev@jackrabbit.apache.org <oa...@jackrabbit.apache.org>
> Subject: Check if a node has children NOT matching a pattern
>
> Hi,
>
> I have a requirement to check if a node has child nodes, which do not match
> a certain pattern. This is definitely important, because
>
> node.hasChildren()
>
> will return true, if there is a "rep:policy" childnode below it. But this
> node is an implementation detail of Oak, and it does not indicate that
> there is a "real" child node below this node. The same happens with
> "jcr:content" child nodes or other implementation specific nodes, which are
> not considered as "independent child nodes" from an application point of
> view.
> The existing API allows only to check for the presence of child nodes
> matching a certain glob (but not "not matching"), so it is of no help for
> me.
>
> Right now it is implemented by iterating over the child nodes and returning
> when the first child is detected which does not match the above criteria.
> This works fine, but it is slow when the child node list is very long (e.g.
> hundreds of nodes). Because then just getting the iterator can take
> miliseconds, especially when the Oak repo is backed by Mongo.
> On top i need to execute this operation quite a number of times within a
> single request, so that getting these iterators can consume quite some
> time, making my requests slower than they should be.
>
> I would like to have an API similar to the existing API, maybe looking like
> this:
>
> node.hasChildren(Pattern notmatchingNodeNames)
>
> which allows me to provide a pattern to it. As soon as a childnode's name
> is found not matching the pattern, the function returns true. This would
> allow the implementation to iterate only over the minimal number of child
> nodes, and I would expect it to be much faster than my existing
> implementation for these larger folders.
>
> I know that the JCR API can hardly be extended by this, but having this in
> a Util class would be fine as well.
>
> WDYT?
>
> Jörg
>
> --
> Cheers,
> Jörg Hoh,
>
> https://cqdump.joerghoh.de
> Twitter: @joerghoh
>
--
Cheers,
Jörg Hoh,
https://cqdump.joerghoh.de
Twitter: @joerghoh
Re: Check if a node has children NOT matching a pattern
Posted by Angela Schreiber <an...@adobe.com.INVALID>.
Hi Jörg
Thanks a lot for reaching out.
IMHO it makes sense to have this reported as feature request for jackrabbit-api as there is no easy way to extend the existing JCR Node API contract.
There already exists a JackrabbitNode interface in oak-jackrabbit-api, which we could use to add the new functionality.
Would you be able to file a feature request in JIRA for this?
Obviously, a patch or PR would be welcome as well.
Kind regards
Angela
PS: i am wondering why the Pattern param would be named 'notmatchingNodeNames'. it could be any kind of regex Pattern, couldn't it?
PS2: for consistency we might also consider adding JackrabbitNode.getChildren(Pattern pattern) alongside the new hasChildren method.
________________________________
From: Jörg Hoh <jh...@googlemail.com.INVALID>
Sent: Friday, January 14, 2022 11:03 AM
To: oak-dev@jackrabbit.apache.org <oa...@jackrabbit.apache.org>
Subject: Check if a node has children NOT matching a pattern
Hi,
I have a requirement to check if a node has child nodes, which do not match
a certain pattern. This is definitely important, because
node.hasChildren()
will return true, if there is a "rep:policy" childnode below it. But this
node is an implementation detail of Oak, and it does not indicate that
there is a "real" child node below this node. The same happens with
"jcr:content" child nodes or other implementation specific nodes, which are
not considered as "independent child nodes" from an application point of
view.
The existing API allows only to check for the presence of child nodes
matching a certain glob (but not "not matching"), so it is of no help for
me.
Right now it is implemented by iterating over the child nodes and returning
when the first child is detected which does not match the above criteria.
This works fine, but it is slow when the child node list is very long (e.g.
hundreds of nodes). Because then just getting the iterator can take
miliseconds, especially when the Oak repo is backed by Mongo.
On top i need to execute this operation quite a number of times within a
single request, so that getting these iterators can consume quite some
time, making my requests slower than they should be.
I would like to have an API similar to the existing API, maybe looking like
this:
node.hasChildren(Pattern notmatchingNodeNames)
which allows me to provide a pattern to it. As soon as a childnode's name
is found not matching the pattern, the function returns true. This would
allow the implementation to iterate only over the minimal number of child
nodes, and I would expect it to be much faster than my existing
implementation for these larger folders.
I know that the JCR API can hardly be extended by this, but having this in
a Util class would be fine as well.
WDYT?
Jörg
--
Cheers,
Jörg Hoh,
https://cqdump.joerghoh.de
Twitter: @joerghoh