Improve UID_IN with variable (done)

ref: Add support of Value Variables for uid_in. · Issue #3066 · dgraph-io/dgraph · GitHub

One of the things that made me think about uid_in needing improviment was the fact that every time we update the Play Dgraph binaries. All UIDs are changed, thus breaking all examples in Docs that use hand-defined UIDs.

So, one idea would be to use the uid_in target as a variable (hand-defined). However, Dgraph does not support rendering UIDs into numbers or strings.

Before continue, here some example of users complaining about uid_in

Continuing

Read my last comment here Add support of Value Variables for uid_in. · Issue #3066 · dgraph-io/dgraph · GitHub

{
  var(func: uid(0x0)){
    a as math(637870) # this will be converted to hex internally.
  }
 
  caro(func: eq(name@en, "Marc Caro")) {
    uid
    name@en
    director.film @filter(uid_in(~director.film, val(a)) ) {
                                # uid_in accepts val()
      name@en
    }
  }

}

I have two ideas to solve this. One simply supports uid_in value variables natively, and the other is adding a feature as “math()” func. But that one would be “hacky”.

This is the desired one

getJeunet as var(func: eq(name@fr, "Jean-Pierre Jeunet")) 

And this example is using the “hacky” way

we get the uid from the var and transform it into a string. To use in uid_in func.

  var(func: eq(name@fr, "Jean-Pierre Jeunet")){
    IN as uid
    a as string(IN)
  }

I’ve mentioned about string() in custom-block

I think we should have uid_in work with query variables. The fewer places we require UID literals, the better. At the very least, it should support the uid() function so you can say uid_in(director.film, uid(varName)).

I just learned last week from @pawan that shortest path queries’s from and to arguments support the uid() function.

This is now supported after this PR: https://github.com/dgraph-io/dgraph/pull/5320 albeit with a uid variable. @MichelDiz can you close this and accept the solution. Also probably edit the heading.

1 Like