Why are there so many
Maybes in my responses? How do I reduce them?
If your GraphQL schema has a nullable field,
dillonkearns/elm-graphql will generate a corresponding
Maybe for that field in the response.
The best way to get rid of these
Maybes is to make the fields non-nullable (for example, turn a
string field into a
string! field or a
[int] into a
If you are unable to change your schema and make the nullable fields non-nullable, you can use
Graphql.SelectionSet.nonNullOrFail. Be aware that the entire response will fail to decode if you do get a null back in a field where you used
What if the
Maybes are inside of a list?
Again, ideally you could change your schema (e.g.
[string!]!). If that's not possible, take a look at the
Graphql.Field.nonNullElementsOrFail function to turn a
List (Maybe something) into a
List something (see the
Take a look at the section about this from the official GraphQL docs.
Why do I get an error when I don't provide an Optional Argument? According to the schema it's optional.
This is very common, if you look at your schema you will probably find that the optional argument is marked as nullable (i.e. it doesn't end with a
!). And in GraphQL, a nullable argument is exactly what an optional argument is, see http://facebook.github.io/graphql/October2016/#sec-Required-Non-Null-Arguments
Arguments can be required. Arguments are required if the type of the argument is non‐null. If it is not non‐null, the argument is optional.
But even though the schema lists the argument as optional (i.e. nullable), it's common to throw a runtime error and say that it's invalid. A common reason for this is that you must pass in "one of the following...", so no individual argument is required, but if you don't provide one it will throw a runtime error. Here's an example from the Github API (you can reporduce it yourself by running this query in the Github Explorer):
Why is the
OptionalArgument union type used instead of
An optional argument can be either present, absent, or null, so using a Maybe does not fully capture the GraphQL concept of an optional argument. For example, you could have a mutation that deletes an entry if a null argument is provided, or does nothing if the argument is absent. See The official GraphQL spec section on null for details.
Where is the best place to store my queries/mutations, at the page/view level or somewhere higher?
How and why does
dillonkearns/elm-graphql change some of the names from my GraphQL schema?
Some GraphQL names are invalid in Elm, and others are not idiomatic and would sound awkard in the context of Elm code.
For example, it is conventional to name a Union's values with all caps in GraphQL, like a union
Episode with values
dillonkearns/elm-graphql will generate the following union type
type Episode = Empire | Jedi | Newhope. If you follow the GraphQL naming conventions,
dillonkearns/elm-graphql will generate nice names that follow Elm naming conventions.
Elm also has to avoid reserved words in the language like, so it would convert a field name like
import_ (See https://github.com/dillonkearns/elm-graphql/pull/41 for more in-depth discussion of this). If you want more details on the normalization behavior you can take a look at the normalization test suite.
What if you have two names that
dillonkearns/elm-graphql normalizes to the same thing, like a field called
User (which would both turn into
user? This is possible, but indicates that you are not following GraphQL conventions. Consider using a different naming convention. If you have a compelling reason for your naming, open an issue so we can discuss the normalization strategy.
How do I upgrade to Elm 0.19?
Take a look at the
dillonkearns/elm-graphql Elm 0.19 upgrade guide.
Wasn't this library called Graphqelm before?
Yes, you can read about why the name changed here.
How do field aliases work in
You may notice that "Show Aliases" is unchecked by default in the demos in the
Aliases are just a tool for telling GraphQL what field to return your data back under
in the JSON response. But with
dillonkearns/elm-graphql, you're not dealing with
JSON directly. You just use
to build up data structures. The JSON response details are all handled under the hood by
Sometimes GraphQL requires you to use aliases to make sure you don't make ambiguous
queries. For example, if you make a request asking for
avatar(size: SMALL) and
avatar(size: LARGE) in the same request, then you will need to use an alias
since they can't both come back under the JSON field
avatar with more than
avatar data in the response.
We hide aliases by default in this demos to reduce noise. But note that these request may be invalid without the aliases, so check "Show Aliases" if you would like to copy-paste the queries and execute them yourself.
You can read more about how aliases are used under the hood in this blog
Why are all my scalars Strings? How do I turn it into other types like Float or Time.Posix?
dillonkearns/elm-graphql deserializes custom scalars into
Strings because scalars are a blindspot in GraphQL. At the moment, there's no way to express what underlying type you use to represent your Scalars in your GraphQL Schema. This is just an area of GraphQL that isn't type-safe yet for some reason. But there is a proposal to allow you (but not require you) to specify the base primitive you use to represent your scalars. See this issue for more on that pending GraphQL Spec change: https://github.com/dillonkearns/elm-graphql/issues/39
As for the type wrappers, like
type Numeric = Numeric String, the reason for that type wrapper is that it allows you to make certain assumptions given a particular Scalar type. If it was just a raw String, then the compiler wouldn't be able to help ensure that if you write a function to transform a particular scalar, you only use it for that scalar and not any other string by mistake.
There are two ways to turn your Custom Scalars into the data types you want.
Use the Custom Scalar Decoders feature of
dillonkearns/elm-graphql. Look at this example and mini-tutorial to see how and what the resulting code looks like.
Or you can unwrap the default scalar wrapper type using
SelectionSet.map. The result looks something like this:
Query.someNumericField |> SelectionSet.map (\(Api.Scalar.Numeric rawNumeric) -> String.toFloat rawNumeric |> Maybe.withDefault 0)
Why is there a custom
Graphql.Http.HttpError type? How do I turn it into
firstname.lastname@example.org introduced a change to the
Http.Error type where
BadStatus went from containing the status code and the message to only containing the status code. Some users need that bit of data (the message for a bad status response). In order to prevent losing information, we needed to define a custom
HttpError type for
elm-graphql. See the full discussion on this issue.
You can use
withSimpleHttpError to turn the data type from a
Graphql.Http.HttpError into a standard
Http.Error. See the example in the docs for
withSimpleHttpError for details on how to use that function.