Can I access the database from a metadata source plugin?

[Jellby]

I know this is not the intended use for a metadata source plugin, but I'd like a plugin to pick up information from a custom column in the database. The identifier I rely on should be unique in the database, so in the "identify" method I can easily select the book record to get information from... if I can access the calibre database. (Any fast and easy way to get a book id from an identifier other than a search, too?)

[kovidgoyal]

You really should not be doing that. It is unsupported and quite likely to break in future versions, but if you really, really want to:

from calibre.gui2.ui import get_gui

db = get_gui().current_db.new_api

Of course, this will break in horrible ways if someone runs the fetch-ebook-metadata command line tool to download metadata without a GUI

[Jellby]

Thanks. This is only intended for internal management of the MR library, so it will not be published. I expect the database access can be tested and gracefully fail if it cannot be found.

[chaley]

Originally Posted by Jellby

(Any fast and easy way to get a book id from an identifier other than a search, too?)

If you are doing many books at one time then I would consider getting all these "identifiers" and storing them in a dict with identifier as key and the book id as data. It might even be faster to do this even if you are doing only two or three books, given that search must scan all the records.

You can get all the values for a column using all_field_for() with book_ids == all_book_ids() then inverting the resulting dict.

[Jellby]

Originally Posted by kovidgoyal

from calibre.gui2.ui import get_gui

db = get_gui().current_db.new_api

I put these inside the "identify" method of the plugin, and when I call it (using the download metadata button in the GUI) I get:

Traceback (most recent call last):
  File "site-packages/calibre/ebooks/metadata/sources/identify.py", line 44, in run
  File "calibre_plugins.dummy1.__init__", line 53, in identify
AttributeError: 'NoneType' object has no attribute 'current_db'

and of course "get_gui()" returns "None".

[kovidgoyal]

Yes, in the identify() method it wont work, since that is run in a worker process. Because of your previous questions, I somewhat sloppily assumed you meant the url_from_identifier() method.

You can just directly read for the database, with something like this,

import apsw
conn = apsw.Connection('/path/to/calibre/metadata.db', apsw.SQLITE_OPEN_READONLY)

apsw is multi-process/thread safe so it should be fine.

[Jellby]

I must hardcode the library path then? I hoped I could get it from the GUI (at least in the cases where the method is triggered from the GUI)

[kovidgoyal]

from calibre.utils.config import prefs
library_path = prefs['library_path']