Using Beautiful Soup from a plugin

[geekraver]

Hi devs

I am working on a plugin for which I want to use Beautiful Soup. However, my plugin fails on the import:

from bs4 import BeautifulSoup

with:

File "calibre_plugins.magazines.bs4.__init__", line 30, in <module>
File "site-packages/calibre/customize/zipplugin.py", line 179, in load_module
File "calibre_plugins.magazines.bs4.builder.__init_ _", line 4, in <module>
ImportError: No module named bs4.element

This is if I bundle bs4 with the plugin. If I don't, and try to rely on the fact that it is installed in my Python distribution, then I get:

File "calibre_plugins.magazines.worker", line 15, in <module>
ImportError: No module named bs4

Has anyone used bs4 in a plugin and have any tips?

Thanks!

[DiapDealer]

1) calibre's python environment already comes with BeautifulSoup, so unless there's an issue with needing a specific version, I'd just import that.

2) Modules included within a plugin's package are only available for import during the "run" method of the plugin (at least that always used to be the case).

2) Import modules using the name of the text-file included in your plugin.

ex: if the filename is "plugin-import-name-my_shiny_plugin.txt" you would use:

from calibre_plugins.my_shiny_plugin import my_shiny_module

3) You can't rely on modules installed in your machine's Python installation, because calibre doesn't use your Python installation.

[geekraver]

Great, thanks! Have it working now.

[geekraver]

Are there any known issues with the Beautiful Soup that is bundled with Calibre? I'm finding that from within my plugin, a soup.find(class_="foo") is returning None, while in my test script hitting the same URL it works and finds the div I want. I logged the page contents and they look fine so everything seems to be working up to the point I call find (I can copy the metadata log to the clipboard and paste it in an editor and see the element that the find failed to locate).

EDIT: I notice that doing finds by id seems to work, but not finds by class.

EDIT: the above was the key. The soup distributed with Calibre is an older version that doesn't support the class_ parameter. Instead you must use:

find(attrs={'class':'foo'})