Add fetchThreadImages

[mlodybercik]

Description

I wanted to create a way of downloading all images posted sent in given thread.

Research

I stareted by watching network traffic on messenger.com while scrolling down the image history.
First query looked like this:
https://www.messenger.com/webgraphql/query/?query_id=515216185516880&variables={"id":"<uid>","first":"<number>"}.Where uid is specified user and number is how many to download, default is 12. Facebook will do more, but I am kinda scared.

If you scroll down further, it shows more and more history.
Query now looks like this:
https://www.messenger.com/webgraphql/query/?query_id=515216185516880&variables={"id":"<uid>","after":"<image_id>","first":<number>}. Previous variables are the same, but there is new thing called after, after digging around some more I found out that image_id is special variable for starting query after given image. It is called coursor in JSON.

After POST'ing it, it returns almost readable JSON, it begins with "for (;;);".

My development

I forked it and created my own crude function to handle it. I'll keep working on it.

[madsmtm]

Hi there @krzesu0, thanks for the research!

I've previously investigated Facebook's GraphQL endpoints quite extensively, and i know that the endpoint at /webgraphql/query works similarly to the one at /api/graphqlbatch/, which we've previously implemented support for, see Client.graphql_request. So I'd suggest using that, something like this should get you far:

data = {"id": <thread_id>, "first": 12}
j = self.graphql_request(_graphql.from_query_id("515216185516880", data))

Note: This internal API was changed recently in #439, so remember to update from master before implementing this!

When you want to parse the result, please make a helper @classmethod on the _file.ImageAttachment class, and convert the result to ImageAttachment objects.

Regarding the cursors, then I think it'd be nice if we could hide them as an implementation detail, and just return an iterable, then the user can choose how many results they want from there?

[mlodybercik]

When you want to parse the result, please make a helper @classmethod on the _file.ImageAttachment class, and convert the result to ImageAttachment objects.

Regarding the cursors, then I think it'd be nice if we could hide them as an implementation detail, and just return an iterable, then the user can choose how many results they want from there?

I rewrote this as @madsmtm suggested. I decided to hide cursor from user . Both _fetchImages (helper) and fetchThreadImages are now generator objects.

[madsmtm]

In general, sorry I caused you so much trouble with the various merges, and thanks for the hard work!

[mlodybercik]

I don't think this feature will work for thread videos, It returns only the preview URL and ID's. It uses these ID's to download it from /mercury/attachments/video/?video_id=.... This could be another feature I'll work on.

[madsmtm]

I tried it out today, and left a few comments, but otherwise good work 👍.

Regarding videos, then I think you're right, it uses another system. You should be fully welcome to implement that (in a separate PR 😉) when you want to!

[madsmtm]

Nice work, I'll go ahead and merge this now! 🎉