Always sort scalar types by name when returning them

Previously these were returned in an essentially random order. This would lead
to flapping in the generated code when running the client CLI against the same
schema multiple times.

In the application I'm writing, I check in the generated schema, so this sort
of flapping in the generated code creates unwanted noise in the commit
history.

Author: autarch

CHANGELOG.md

@@ -12,6 +12,9 @@ and this project adheres to [Semantic Versioning](http://semver.org/spec/v2.0.0.
   `rustls` rather than `native-tls`.
 - Code generated by the `graphql-client` CLI program now suppresses all
   warnings from rustc and clippy.
+- The generated code now sorts scalar types by name. This fixes "flapping" in
+  the generated code that used to happen as the scalar types would be randomly
+  ordered each time the generator ran.
 
 ## 0.10.0 - 2021-07-04
 

graphql_client_codegen/Cargo.toml

@@ -8,6 +8,7 @@ repository = "https://github.com/graphql-rust/graphql-client"
 edition = "2018"
 
 [dependencies]
+itertools = "0.10.3"
 graphql-introspection-query = { version = "0.2.0", path = "../graphql-introspection-query" }
 graphql-parser = "^0.2"
 heck = "0.4.0"

graphql_client_codegen/src/query.rs

@@ -18,6 +18,7 @@ use crate::{
         StoredInputType, StoredScalar, TypeId, UnionId,
     },
 };
+use itertools::Itertools;
 use std::{
     collections::{HashMap, HashSet},
     fmt::Display,
@@ -605,6 +606,7 @@ impl UsedTypes {
             .filter_map(TypeId::as_scalar_id)
             .map(move |scalar_id| (scalar_id, schema.get_scalar(scalar_id)))
             .filter(|(_id, scalar)| !crate::schema::DEFAULT_SCALARS.contains(&scalar.name.as_str()))
+            .sorted_by_key(|(_id, scalar)| scalar.name.as_str())
     }
 
     pub(crate) fn enums<'a, 'schema: 'a>(