On Wed, Aug 20, 2014 at 4:08 PM, Norbert Preining <[email protected]>
wrote:
In the first step I would implement only *query* support, so no
syncronization with a running shotwell is necessary.
Doing operations via the cmd line is more advanced and I am not
sure if that makes sense.
Yes, I would stick with queries to begin with. To give a broad
overview of Shotwell's database-memory model:
* Shotwell *heavily* caches the database in memory for speed reasons.
Thus, a query for information is probably best served by looking at the
objects in memory and returning that rather than an actual SQLite
transaction.
* Shotwell only goes out to the database to read information when it
does not have it in memory. There's very little no kept in memory,
however.
* Shotwell writes changes to the database as soon as the user requests
it. That is, Shotwell doesn't delay the write in order to do it in the
background, a common scheme in these kinds of situations.
I guess the question I have is, what kind of syntax would this CLI
have. One approach would be straight-up SQL syntax, but I would
strongly argue against that; the whole point of this is that the
database schema should not be exposed to users because it can change at
any time. Just to get the ball rolling, I can imagine a dot schema
that works something like this ("shq" is "Shotwell query"):
$ swq event[12345].description event[12345].title event[12345].photo_ids
Returns:
Beautiful Tokyo
Tokyo Trip, 2014
101;102;103;104;110;112;
The command-line accepts attributes that can be queried. The number in
the square brackets is the event identifier. event.description and
event.name return a string. (We will need to deal with multiline
strings, as description supports them.) event.photo_ids generates a
semicolon-delimited list of photo identifiers. From this a program
could then execute:
$ swq photo[101].title photo[102].camera photo[103].orientation
... and so forth. Note that photo.camera is referring to the EXIF
camera identifier, which is not stored in the database or memory. In
this situation, Shotwell would read the metadata from the photo file
and return that to the caller.
I'm just riffing off the top of my head, but this is the direction I'm
thinking of.
Anyway, with this said, I am happy to dig (again) into some shotwell
features and bugs ;-) One is actually open since long, the comment
editing of events feom the event screent jumps to the comment of
the first photo instead of the event comment ;-)
Great! Look forward to seeing it.
-- Jim
_______________________________________________
shotwell-list mailing list
[email protected]
https://mail.gnome.org/mailman/listinfo/shotwell-list
