For Loops​

Structured params​

const query = e.params(
  {
    characters: e.array(
      e.tuple({
        portrayed_by: e.str,
        name: e.str,
        movies: e.array(e.str),
      })
    ),
  },
  (params) => { ...

In raw EdgeQL, you can only have scalar types as parameters. We could mirror that here with something like this: e.params({characters: e.json}), but this would then require us to cast all the values inside the JSON like portrayed_by and name.

By doing it this way — typing characters with e.array and the character objects as named tuples by passing an object to e.tuple — all the data in the array will be properly cast for us. It will also better type check the data you pass to the query’s run method.

Inserting the inner conflicting data​

...
(params) => {
  const movies = e.for(
    e.op("distinct", e.array_unpack(e.array_unpack(params.characters).movies)),
    (movie) => {
      return e
        .insert(e.Movie, {
          title: movie,
        })
        .unlessConflict((movie) => ({
          on: movie.title,
          else: movie,
        }));
    }
  );
...

We need to separate this movie insert query so that we can use distinct on it. We could just nest an insert inside our character insert if movies weren’t duplicated across characters (e.g., two characters have “The Avengers” in movies). Even though the query is separated from the character inserts here, it will still be built as part of a single EdgeDB query using with which we’ll get to a bit later.

The distinct operator can only operate on sets. We use array_unpack to make these arrays into sets. We need to call it twice because params.characters is an array and .movies is an array nested inside each character.

Chaining unlessConflict takes care of any movies that already exist in the database before we run this query, but it won’t handle conflicts that come about over the course of this query. The distinct operator we used earlier pro-actively eliminates any conflicts we might have had among this data.

Inserting the outer data​

...
return e.with(
  [movies],
  e.for(e.array_unpack(params.characters), (character) => {
    return e.insert(e.Character, {
      name: character.name,
      portrayed_by: character.portrayed_by,
      movies: e.assert_distinct(
        e.select(movies, (movie) => ({
          filter: e.op(movie.title, "in", e.array_unpack(character.movies)),
        }))
      ),
    });
  })
);
...

The query builder will try to automatically use EdgeQL’s with, but in this instance, it doesn’t know where to place the with. By using e.with explicitly, we break our movie insert out to the top-level of the query. By default, it would be scoped inside the query, so our distinct operator would be applied only to each character’s movies instead of to all of the movies. This would have caused the query to fail.

The rest of the query is relatively straightforward. We unpack params.characters to a set so that we can pass it to e.for to iterate over the characters. For each character, we build an insert query with their name and portrayed_by values.

For the character’s movies, we select everything in the movies insert query we wrote previously, filtering for those with titles that match values in the character.movies array.

All that’s left is to run the query, passing the data to the query’s run method!

const query = e.params(
  {
    characters: e.array(
      e.tuple({
        name: e.str,
        last_appeared: e.int64,
      })
    ),
  },
  (params) => {
    return e.for(e.array_unpack(params.characters), (character) => {
      return e.update(e.Character, () => ({
        filter_single: { name: character.name },
        set: {
          last_appeared: character.last_appeared,
        },
      }));
    });
  }
);

await query.run(client, {
  characters: [
    { name: "Iron Man", last_appeared: 2019 },
    { name: "Captain America", last_appeared: 2019 },
    { name: "The Hulk", last_appeared: 2021 },
  ],
});