Firefox bookmarks plugin in Python?

[tomsquest]

Hi,

With the deprecated Firefox bookmarks plugin, I was wondering if porting the old plugin to a Python plugin would work?

Do you think it is technically possible, @ManuelSchneid3r ?

Along, I was wondering if Albert has a periodical refresh of entries? (eg. I add a bookmark to Firefox, when this new entry will appear in Albert).

Thanks

[ManuelSchneid3r]

sure techincally it is possible, but id rather implement it in c++ though. some thoughts:

• filewatches are needed, Qt has them, platform agnostic. in python you would have to introduce dependencies
• sqlite: qt has it and it is loaded anyway by the core. in python you would have to introduce dependencies
• indexing: in python this would lock the GIL, which blocks queries from python query handlers. in c++ you could use threads or the background executor which has a lean interface (example).
• config: the python plugin config widget interface is limited. native plugins have the full widgets set.

The Python plugin interface is nice to have as we can write custom plugins in like minutes, but for sophisticated plugins I'd generally rather recommend native plugins. Therefore I try to make the C++ API as lean and easy to understand as possible. Also there are some features on my todolist (items details view, items drag'n'drop) that will never find their way into the python API.

[tomsquest]

Top, so basically this boils down to c++ for power, and python for convenience.

Thanks for your answer.

[ManuelSchneid3r]

Closing this. Its on the wishlist.

[stevenxxiu]

@tomsquest I made https://github.com/stevenxxiu/albert_firefox in case you're interested.

[tomsquest]

Awesome @stevenxxiu ! 💯
I will install it ASAP. This is a must for productivity!
I will report any issue to this repo, Steven.

[ManuelSchneid3r]

Is anybody interested in porting this to C++? I'd give assistance. I have no time to dig into the database design (I know myself, I'd not stop before it's perfect) but a lot of users would benefit from a polished implementation. I see several points that could be improved from here on

• matching is done manually in triggerqueryhandler
• indexqueryhandler does matching implicitly and provides a global handler
• no file watches, has to be updated manually
• locks are not handled. I think the database should be copied before read.

[stevenxxiu]

Have a look at https://albertlauncher.github.io/reference/classalbert_1_1IndexQueryHandler.html and https://albertlauncher.github.io/reference/classalbert_1_1Matcher.html.

Ah I meant some sort of fuzzy matching, but I see you do have that there. Another issue here though is that I always want to show matches for bookmark names before matches for bookmark URLs. Is there some option for that?

fair. any other ideas? do you know of any ideas like table watches or insert hooks or something?

I'm not sure. I can't even open the database without copying it.

[ManuelSchneid3r]

i'd not use urls at all. maybe the tld, but thats it. urls are not meant to be display strings and often contain hashes and other cryptic stuff.

Ah I meant some sort of fuzzy matching

also normalization/diactrics, case and word order.

[ManuelSchneid3r]

If you really want to have that your best bet is probably using matcher and adjusting the score such that title matches get score/2+0.5 and urls just score/2

[ManuelSchneid3r]

but then again this would screw up the interleaved order with other items. if you use matcher, you probably get some kind of what you want because an url is usually long which leads to low scores (len(matched string)/len(entire string)). you also want to not show items twice so i guess a simple max(match(title), match(url)) per item yields the best UX. btw the indexer does such things implicitly.
see how chromium plugin does it.

[stevenxxiu]

Thanks I forgot about the Matcher API. I suppose I didn't need to reimplement something similar then. I switched to that.