Re: FileSystemStorage / CMFEditions integration

On Friday 16 June à 12:05, Encolpe Degoute wrote:
> Sylvain Thénault a écrit :
> > Hi there !
> > 
> > This morning I took the time to extract proper patches that I had have
> > to write to have FileSystemStorage fields working properly with
> > CMFEditions versioning. For this I had to patch:
> > 
> > * CMFEditions, to:
> >   * give a rollback flag to afterRetrieveModifier, allowing to know
> >     in a modifier if an object is retieved for a rollback or not (simple
> >     view of an old version)
> >   * handle a .retrieved_version_id attribute on versioned object to know
> >     the cmfedition version number of an object (at the time at wrote the
> >     patch, I didn't find any attribute giving this information as I'ld
> >     have expected, version_id wasn't suiting my need)
> > 
> > * FileSystemStorage, to:
> >   * handle version number in generated file name
> >   * store fssinfo into different attribute according to object's version
> >     (I don't know enough about cmfeditions'internals to explain that,
> >     but I had have to do this to properly deals with file deletion and
> >     maybe other problems i don't remember right now)
> > 
> > * Archetypes'file widget macro to include version number into its
> >   download link
> > 
> > and finally to write a modifier, dealing with filesystem files on
> > version creation/rollback.
> > 
> > As you can see, the problem is that it introduces some cmfeditions
> > related code into FSS and Archetypes. However this is in restricted
> > place and doesn't cause any problem if cmfeditions is not used or even 
> > installed, and this integration seems to me potentially interesting for
> > others, so I wish to integrate it into the main stream of those 3
> > products. Is someone object on this or have any comment regarding the
> > patches? (see attached files) 
> 
> Hi,
> Like previous patch you give there you add features but there's no unit tests
> with them.

I know. I've reextracted those patch from some client code since I've
upgraded dependencies, but I didn't change those nor write unittest
since the last time I submit them. If there is a consensus to integrate
this into respective products, I'll take the time to write them before
integration.

> Around FSS patch, you hardcode relation between CMFEdition and FSS but
> have you test the impact of your patch with ArrayField and CompoundField ?

no, i'wasn't aware about those field types...

-- 
Sylvain Thénault                               LOGILAB, Paris (France)
Formations Python, Zope, Plone, Debian:  http://www.logilab.fr/formations
Développement logiciel sur mesure:       http://www.logilab.fr/services
Python et calcul scientifique:           http://www.logilab.fr/science