Fixed move bug

[dantleech]

The UUID of a node is indexed by the nodes path. This commit fixes
an issue where this index is not removed, making the transport believe
that there is still a node existing at this path.

[wjzijderveld]

I found the same bug: #158
I also created a PR to test this with the phpcr-api-tests: phpcr/phpcr-api-tests#122

[dbu]

this has the same issue as #158 i guess, it does not handle moved children. cc20a28

i am kind of surprised why we have a path to uuid cache in the transport AND a uuid to path cache in the ObjectManager. do we really need both?

[dbu]

i merged #165 - @dantleech can you add a test about a child of the moved node and then call the method i added to transport to invalidate the cache and see if that fixes the problem?

[dbu]

this might be related to jackalope/jackalope#191

[dbu]

@dantleech can you check if jackalope/jackalope#200 matters for this or not?

[lsmith77]

needs a rebase

[dantleech]

Updated. I added a test for the child - but something very strange happens, as in the test case:

• Create a node called /topic1
• Add a child to that node called /topic1child
• Move /topic1 to /topic2
• Save the session

Now the following path is in the database: /topic2/topic2child -- havn't got time to investigate further today.

[lsmith77]

maybe this is a bug in how we replace?

see https://github.com/dantleech/jackalope-doctrine-dbal/blob/f3bcfaa7caade9fbce4355f471315d3c3dcb6361/src/Jackalope/Transport/DoctrineDBAL/Client.php#L1482

[lsmith77]

see also http://stackoverflow.com/questions/1252693/php-str-replace-that-only-acts-on-the-first-match

[dbu]

ah, interesting. anybody on it right now or shall i have a look?

[dantleech]

I'm working all day so feel free :)

[dbu]

continued in #168 - thanks lukas, your tip made it simple to fix.

[dbu]

btw, i quickly checked if we might have a similar problem in phpcr-odm, but we only use str_replace in places where it makes sense.