jsonpath: mix two paths - facebook-graph-api

Question

I am querying facebook for fullsize profilepictures of users that commented on posted pictures or who like posted pictures by the /me user, using the graph api.
I am trying to either:
do it in one query, where i am having trouble getting data from deep enough.
or minimize the amount of subqueries
In one query: this does not work (it does not give details of the users that commented):
me/photos/uploaded?fields=from,images,comments.picture,likes.picture&limit=500
and specifying the dimensions throws an error:
me/photos/uploaded?fields=from,images,comments.picture,likes.picture.type(large)&limit=500
There seems to be a limit as to how deep you can specify, though I havent found anything explicit in the documentation. (which is so terse that is hardly a big surprise)
so, resolving to batched queries with a dependency
myphotos => me/photos/uploaded?fields=created_time,name,id,from,comments.summary(1).limit(100),images,likes.summary(1).limit(100)&limit=500
userinfo_photocomments => ?ids={result=myphotos:$.data.*.comments.data.*.from.id}&fields=id,name,updated_time,picture.redirect(0).type(large).height(2000).width(2000)
userinfo_photolikes => ?ids={result=myphotos:$.data.*.likes.data.*.id}&fields=id,name,updated_time,picture.redirect(0).type(large).height(2000).width(2000)
though this works, the fact that i cant find how to do it in one or max 2 queries annoys me, especially as the results of the last 2 subqueries will most likely overlap.
I would like to find a way to mix the jsonpath anchors, such that i can combine the last two queries.
The problem seems to be that the userid's of the commentors and the likers are on different depths in the trees.
I would be gratefull for any insights.

Answer 1

I think you should be able to use the result from the first query twice, resulting in only two queries.
I tested the ?ids= endpoint if it filters out duplicate user_id like the following:
GET /?ids={user_id1},{user_id2},{user_id1}
which results in only getting two distinct user_ids back.
So, this yields in the following batch query:
curl \
-F 'access_token={access_token}' \
-F 'batch=[{ "method":"GET","name":"myphotos","relative_url":"me/photos/uploaded?fields=created_time,name,id,from,comments.summary(1).limit(100),images,likes.summary(1).limit(100)&limit=500",},{"method":"GET","relative_url":"?ids={result=myphotos:$.data.*.comments.data.*.from.id},{result=myphotos:$.data.*.likes.data.*.id}&fields=id,name,updated_time,picture.redirect(0).type(large).height(2000).width(2000)"}]' \
https://graph.facebook.com/
Just replace {access_token} with an actual access token to test this.
Basically, I just concatenated the user_id lists for comments and likes as
GET ?ids={result=myphotos:$.data.*.comments.data.*.from.id},{result=myphotos:$.data.*.likes.data.*.id}&fields=...