problems with LibraryDatabase2

[capink]

I want to update the find duplicates plugin to use get_proxy_metadata instead of get_metadata, as per chaley's advice here. There is no problem doing this when comparing duplicates in one library. However problems arise when trying to find cross library duplicates. The original code by kiwidude initializes the target library as follows:

from calibre.library.database2 import LibraryDatabase2
self.target_db = LibraryDatabase2(self.library_path, read_only=True, is_second_db=True)

problem with the above approach is that I cannot use new_api methods like get_proxy_metadata. I get the following error:

AttributeError: 'LibraryDatabase2' object has no attribute 'new_api'

My alternative approach would be initializing the target_db as follows:

from calibre.library import db as DB
self.target_db = DB(self.library_path, read_only=True)

Is there any reason to avoid using the second approach? I am having doubts because this seems like the obvious choice, so maybe kiwidude deliberately avoided that for a reason not obvious to me.

[chaley]

LibraryDatabase2 is obsolete. It works with the new database introduced in V4 because Kovid supplied a "legacy" replacement.

For all new code it is best to use something like this:

def init_cache(library_path):
   from calibre.db.backend import DB
   from calibre.db.cache import Cache
   backend = DB(library_path)  # other named parameters as needed
   cache = Cache(backend)
   cache.init()
   return cache

Use the methods in Cache.

For older code that uses methods in LibraryDatabase2, use something like this:

from calibre.db.legacy import LibraryDatabase
db = LibraryDatabase(library_path) # other named parameters as needed

In this case the methods in "Cache" are available via db.new_api.

I think the new_api attribute might be available in LibraryDatabase2 through some sort of monkeypatching, but I am not sure of that.

[chaley]

Originally Posted by capink

Is there any reason to avoid using the second approach? I am having doubts because this seems like the obvious choice, so maybe kiwidude deliberately avoided that for a reason not obvious to me.

This didn't exist when Kiwidude wrote the plugin.

[capink]

Thank you for the reply. What does the extra Cache(backend) do exactly? What would be the difference if I skipped it?

[chaley]

Originally Posted by capink

Thank you for the reply. What does the extra Cache(backend) do exactly? What would be the difference if I skipped it?

Cache contains the API you use, for example get_proxy_metadata. You very seldom use anything in Backend.

[chaley]

Originally Posted by capink

from calibre.library import db as DB
self.target_db = DB(self.library_path, read_only=True)

BTW: this is the same as my second example.

[capink]

OK, got it. Thanks.

[kovidgoyal]

The canonical way to do this is:

from calibre.library import db
db = db('path/to/library/folder').new_api

[capink]

Thanks, Kovid. Will do it that way.

• What is the earliest calibre version that supports this?
• Is there any use for the older interface? the one you get from gui.current_db, or should I always use the new_api? I want to replace all db references with the new_api if possible. If not possible I will have to use a naming convention to differentiate them to avoid confusion.
• when calibre passes references to db — e.g. library_changed(self, db) — which one does it reference?

[kovidgoyal]

Lord knows, probably 2.0, possibly 1.0. No there is never a need to use the old interface in new code. It doesnt matter, you can alway use new_api, regardless of which interface it is. You can for instance do: db.new_api.new_api.new_api ad infinitum.

[capink]

Thanks.