Separate business and exchange types (#419)

### What

We want to be able to confidently evolve the configuration format
without risk of accidentally changing the behavior of previous versions.

This PR has no user-observable change in behavior.

Introducing a new major version of the configuration format is now a
much more well-defined task, which should consist most of just
duplicating the entire `version3` module hierarchy and then introducing
the new changes in the duplicated copy.

### How

In order to achieve this we duplicate the complete metadata type
hierarchy into configuration crate and introduce functions that convert
between the now separate exchange types and business types, and ensure
that all logic that has to do with serialization only lives with the
exchange types (the version types)

crates/configuration/src/configuration.rs

@@ -4,13 +4,12 @@ use std::path::Path;
 
 use schemars::JsonSchema;
 use serde::{Deserialize, Serialize};
-use tokio::fs;
 
 use query_engine_metadata::metadata;
 
 use crate::environment::Environment;
 use crate::error::Error;
-use crate::values::{ConnectionUri, IsolationLevel, PoolSettings, Secret};
+use crate::values::{IsolationLevel, PoolSettings};
 use crate::version3;
 
 pub const CONFIGURATION_FILENAME: &str = "configuration.json";
@@ -59,45 +58,6 @@ pub async fn parse_configuration(
     configuration_dir: impl AsRef<Path>,
     environment: impl Environment,
 ) -> Result<Configuration, Error> {
-    let configuration_file = configuration_dir.as_ref().join(CONFIGURATION_FILENAME);
-
-    let configuration_file_contents =
-        fs::read_to_string(&configuration_file)
-            .await
-            .map_err(|err| {
-                Error::IoErrorButStringified(format!("{}: {}", &configuration_file.display(), err))
-            })?;
-    let mut configuration: version3::RawConfiguration =
-        serde_json::from_str(&configuration_file_contents).map_err(|error| Error::ParseError {
-            file_path: configuration_file.clone(),
-            line: error.line(),
-            column: error.column(),
-            message: error.to_string(),
-        })?;
-    // look for native query sql file references and read from disk.
-    for native_query_sql in configuration.metadata.native_queries.0.values_mut() {
-        native_query_sql.sql = metadata::NativeQuerySqlEither::NativeQuerySql(
-            native_query_sql
-                .sql
-                .from_external(configuration_dir.as_ref())
-                .map_err(Error::IoErrorButStringified)?,
-        );
-    }
-    let connection_uri =
-        match configuration.connection_settings.connection_uri {
-            ConnectionUri(Secret::Plain(uri)) => Ok(uri),
-            ConnectionUri(Secret::FromEnvironment { variable }) => environment
-                .read(&variable)
-                .map_err(|error| Error::MissingEnvironmentVariable {
-                    file_path: configuration_file,
-                    message: error.to_string(),
-                }),
-        }?;
-    Ok(Configuration {
-        metadata: configuration.metadata,
-        pool_settings: configuration.connection_settings.pool_settings,
-        connection_uri,
-        isolation_level: configuration.connection_settings.isolation_level,
-        mutations_version: configuration.mutations_version,
-    })
+    // Try parsing each supported version in turn
+    version3::parse_configuration(configuration_dir, environment).await
 }

crates/configuration/src/version3/comparison.rs

@@ -1,6 +1,6 @@
 //! Helpers for the comparison operators configuration.
 
-use query_engine_metadata::metadata::database::OperatorKind;
+use super::database::OperatorKind;
 use schemars::JsonSchema;
 use serde::{Deserialize, Serialize};
 

crates/configuration/src/version3/metadata/database.rs

@@ -0,0 +1,295 @@
+//! Metadata information regarding the database and tracked information.
+
+// This code was copied from a different place that predated the introduction of clippy to the
+// project. Therefore we disregard certain clippy lints:
+#![allow(
+    clippy::enum_variant_names,
+    clippy::upper_case_acronyms,
+    clippy::wrong_self_convention
+)]
+use schemars::JsonSchema;
+use serde::{Deserialize, Serialize};
+use std::collections::{BTreeMap, BTreeSet};
+
+/// Map of all known composite types.
+#[derive(Debug, Clone, PartialEq, Eq, Default, Serialize, Deserialize, JsonSchema)]
+#[serde(rename_all = "camelCase")]
+pub struct CompositeTypes(pub BTreeMap<String, CompositeType>);
+
+/// A Scalar Type.
+#[derive(Debug, Clone, PartialEq, Eq, PartialOrd, Ord, Serialize, Deserialize, JsonSchema)]
+#[serde(rename_all = "camelCase")]
+pub struct ScalarType(pub String);
+
+/// The type of values that a column, field, or argument may take.
+#[derive(Debug, Clone, PartialEq, Eq, Deserialize, Serialize, JsonSchema)]
+#[serde(rename_all = "camelCase")]
+pub enum Type {
+    ScalarType(ScalarType),
+    CompositeType(String),
+    ArrayType(Box<Type>),
+}
+
+/// Information about a composite type. These are very similar to tables, but with the crucial
+/// difference that composite types do not support constraints (such as NOT NULL).
+#[derive(Debug, Clone, PartialEq, Eq, Serialize, Deserialize, JsonSchema)]
+#[serde(rename_all = "camelCase")]
+pub struct CompositeType {
+    pub name: String,
+    pub fields: BTreeMap<String, FieldInfo>,
+    #[serde(default)]
+    pub description: Option<String>,
+}
+
+/// Information about a composite type field.
+#[derive(Debug, Clone, PartialEq, Eq, Serialize, Deserialize, JsonSchema)]
+#[serde(rename_all = "camelCase")]
+pub struct FieldInfo {
+    pub name: String,
+    pub r#type: Type,
+    #[serde(default)]
+    pub description: Option<String>,
+}
+
+/// The complete list of supported binary operators for scalar types.
+/// Not all of these are supported for every type.
+#[derive(Debug, Clone, PartialEq, Eq, Default, Serialize, Deserialize, JsonSchema)]
+#[serde(rename_all = "camelCase")]
+pub struct ComparisonOperators(pub BTreeMap<ScalarType, BTreeMap<String, ComparisonOperator>>);
+
+/// Represents a postgres binary comparison operator
+#[derive(Debug, Clone, PartialEq, Eq, Serialize, Deserialize, JsonSchema)]
+#[serde(rename_all = "camelCase")]
+pub struct ComparisonOperator {
+    pub operator_name: String,
+    pub operator_kind: OperatorKind,
+    pub argument_type: ScalarType,
+
+    #[serde(default = "default_true")]
+    pub is_infix: bool,
+}
+
+/// Is it a built-in operator, or a custom operator.
+#[derive(Debug, Clone, PartialEq, Eq, Serialize, Deserialize, JsonSchema)]
+#[serde(rename_all = "camelCase")]
+pub enum OperatorKind {
+    Equal,
+    In,
+    Custom,
+}
+
+/// This is quite unfortunate: https://github.com/serde-rs/serde/issues/368
+/// TL;DR: we can't set default literals for serde, so if we want 'is_infix' to
+/// default to 'true', we have to set its default as a function that returns 'true'.
+fn default_true() -> bool {
+    true
+}
+
+/// Mapping from a "table" name to its information.
+#[derive(Debug, Clone, PartialEq, Eq, Default, Serialize, Deserialize, JsonSchema)]
+#[serde(rename_all = "camelCase")]
+pub struct TablesInfo(pub BTreeMap<String, TableInfo>);
+
+/// Information about a database table (or any other kind of relation).
+#[derive(Debug, Clone, PartialEq, Eq, Serialize, Deserialize, JsonSchema)]
+#[serde(rename_all = "camelCase")]
+pub struct TableInfo {
+    pub schema_name: String,
+    pub table_name: String,
+    pub columns: BTreeMap<String, ColumnInfo>,
+    #[serde(default)]
+    pub uniqueness_constraints: UniquenessConstraints,
+    #[serde(default)]
+    pub foreign_relations: ForeignRelations,
+    #[serde(default)]
+    pub description: Option<String>,
+}
+
+/// Can this column contain null values
+#[derive(Debug, Clone, PartialEq, Eq, Default, Serialize, Deserialize, JsonSchema)]
+#[serde(rename_all = "camelCase")]
+pub enum Nullable {
+    #[default]
+    Nullable,
+    NonNullable,
+}
+
+/// Does this column have a default value.
+#[derive(Debug, Clone, PartialEq, Eq, Default, Serialize, Deserialize, JsonSchema)]
+#[serde(rename_all = "camelCase")]
+pub enum HasDefault {
+    #[default]
+    NoDefault,
+    HasDefault,
+}
+
+/// Is this column an identity column.
+#[derive(Debug, Clone, PartialEq, Eq, Default, Serialize, Deserialize, JsonSchema)]
+#[serde(rename_all = "camelCase")]
+pub enum IsIdentity {
+    #[default]
+    NotIdentity,
+    IdentityByDefault,
+    IdentityAlways,
+}
+
+/// Is this column a generated column.
+#[derive(Debug, Clone, PartialEq, Eq, Default, Serialize, Deserialize, JsonSchema)]
+#[serde(rename_all = "camelCase")]
+pub enum IsGenerated {
+    #[default]
+    NotGenerated,
+    Stored,
+}
+
+/// Information about a database column.
+#[derive(Debug, Clone, PartialEq, Eq, Serialize, Deserialize, JsonSchema)]
+#[serde(rename_all = "camelCase")]
+pub struct ColumnInfo {
+    pub name: String,
+    pub r#type: Type,
+    #[serde(default)]
+    pub nullable: Nullable,
+    #[serde(skip_serializing_if = "does_not_have_default")]
+    #[serde(default)]
+    pub has_default: HasDefault,
+    #[serde(skip_serializing_if = "is_not_identity")]
+    #[serde(default)]
+    pub is_identity: IsIdentity,
+    #[serde(skip_serializing_if = "is_not_generated")]
+    #[serde(default)]
+    pub is_generated: IsGenerated,
+    #[serde(default)]
+    pub description: Option<String>,
+}
+
+fn does_not_have_default(has_default: &HasDefault) -> bool {
+    matches!(has_default, HasDefault::NoDefault)
+}
+
+fn is_not_identity(is_identity: &IsIdentity) -> bool {
+    matches!(is_identity, IsIdentity::NotIdentity)
+}
+
+fn is_not_generated(is_generated: &IsGenerated) -> bool {
+    matches!(is_generated, IsGenerated::NotGenerated)
+}
+
+/// A mapping from the name of a unique constraint to its value.
+#[derive(Debug, Clone, PartialEq, Eq, Default, Serialize, Deserialize, JsonSchema)]
+#[serde(rename_all = "camelCase")]
+pub struct UniquenessConstraints(pub BTreeMap<String, UniquenessConstraint>);
+
+/// The set of columns that make up a uniqueness constraint.
+#[derive(Debug, Clone, PartialEq, Eq, Serialize, Deserialize, JsonSchema)]
+#[serde(rename_all = "camelCase")]
+pub struct UniquenessConstraint(pub BTreeSet<String>);
+
+/// A mapping from the name of a foreign key constraint to its value.
+#[derive(Debug, Clone, PartialEq, Eq, Default, Serialize, Deserialize, JsonSchema)]
+#[serde(rename_all = "camelCase")]
+pub struct ForeignRelations(pub BTreeMap<String, ForeignRelation>);
+
+/// A foreign key constraint.
+#[derive(Debug, Clone, PartialEq, Eq, Serialize, Deserialize, JsonSchema)]
+#[serde(rename_all = "camelCase")]
+pub struct ForeignRelation {
+    #[serde(skip_serializing_if = "Option::is_none")]
+    pub foreign_schema: Option<String>,
+    pub foreign_table: String,
+    pub column_mapping: BTreeMap<String, String>,
+}
+
+/// All supported aggregate functions, grouped by type.
+#[derive(Debug, Clone, PartialEq, Eq, Default, Serialize, Deserialize, JsonSchema)]
+#[serde(rename_all = "camelCase")]
+pub struct AggregateFunctions(pub BTreeMap<ScalarType, BTreeMap<String, AggregateFunction>>);
+
+#[derive(Debug, Clone, PartialEq, Eq, Serialize, Deserialize, JsonSchema)]
+#[serde(rename_all = "camelCase")]
+pub struct AggregateFunction {
+    pub return_type: ScalarType,
+}
+
+/// Type representation of scalar types, grouped by type.
+#[derive(Debug, Clone, PartialEq, Eq, Default, Serialize, Deserialize, JsonSchema)]
+#[serde(rename_all = "camelCase")]
+pub struct TypeRepresentations(pub BTreeMap<ScalarType, TypeRepresentation>);
+
+/// Type representation of a scalar type.
+#[derive(Debug, Clone, PartialEq, Eq, Serialize, Deserialize, JsonSchema)]
+#[serde(rename_all = "camelCase")]
+pub enum TypeRepresentation {
+    /// JSON booleans
+    Boolean,
+    /// Any JSON string
+    String,
+    /// float4
+    Float32,
+    /// float8
+    Float64,
+    /// int2
+    Int16,
+    /// int4
+    Int32,
+    /// int8
+    Int64,
+    /// numeric
+    BigDecimal,
+    /// timestamp
+    Timestamp,
+    /// timestamp with timezone
+    Timestamptz,
+    /// time
+    Time,
+    /// time with timezone
+    Timetz,
+    /// date
+    Date,
+    /// uuid
+    UUID,
+    /// geography
+    Geography,
+    /// geometry
+    Geometry,
+    /// Any JSON number
+    Number,
+    /// Any JSON number, with no decimal part
+    Integer,
+    /// An arbitrary json.
+    Json,
+    /// One of the specified string values
+    Enum(Vec<String>),
+}
+
+// tests
+
+#[cfg(test)]
+mod tests {
+    use super::{ScalarType, TypeRepresentation, TypeRepresentations};
+
+    #[test]
+    fn parse_type_representations() {
+        assert_eq!(
+            serde_json::from_str::<TypeRepresentations>(
+                r#"{"int4": "integer", "card_suit": {"enum": ["hearts", "clubs", "diamonds", "spades"]}}"#
+            )
+            .unwrap(),
+            TypeRepresentations(
+                [(
+                    ScalarType("int4".to_string()),
+                    TypeRepresentation::Integer
+                ), (
+                    ScalarType("card_suit".to_string()),
+                    TypeRepresentation::Enum(vec![
+                        "hearts".into(),
+                        "clubs".into(),
+                        "diamonds".into(),
+                        "spades".into()
+                    ])
+                )]
+                .into()
+            )
+        );
+    }
+}