feat: CLI command to autogenerate JSON Schema for PL, RQ and lineage (#…

…4698)

CHANGELOG.md

@@ -6,6 +6,10 @@
 
 **Features**:
 
+- Added `prqlc debug json-schema` command to auto-generate JSON Schema
+  representations of commonly exposed IR types such as PL and RQ. (@kgutwin,
+  #4698)
+
 **Fixes**:
 
 - Using `in` with an empty array pattern (e.g. `expr | in []`) will now output a

Cargo.toml

@@ -51,6 +51,7 @@ insta = {version = "1.39.0", features = ["colors", "glob", "yaml"]}
 insta-cmd = "0.6.0"
 itertools = "0.13.0"
 log = "0.4.22"
+schemars = "1.0.0-alpha.2"
 semver = {version = "1.0.23", features = ["serde"]}
 serde = {version = "1.0.203", features = ["derive"]}
 serde_json = "1.0.120"

prqlc/prqlc-parser/Cargo.toml

@@ -16,6 +16,7 @@ doctest = false
 enum-as-inner = {workspace = true}
 itertools = {workspace = true}
 log = {workspace = true}
+schemars = {workspace = true}
 semver = {version = "1.0.23", features = ["serde"]}
 serde = {workspace = true}
 serde_yaml = {workspace = true, optional = true}

prqlc/prqlc-parser/src/generic.rs

@@ -4,11 +4,12 @@
 // hasn't been used much since, and I'm not sure carries its weight. So we
 // could consider rolling back to only concrete implementations to delayer the
 // code.
+use schemars::JsonSchema;
 use serde::{Deserialize, Serialize};
 
 /// Inclusive-inclusive range.
 /// Missing bound means unbounded range.
-#[derive(Debug, Clone, Default, Serialize, Deserialize, PartialEq, Eq)]
+#[derive(Debug, Clone, Default, Serialize, Deserialize, PartialEq, Eq, JsonSchema)]
 pub struct Range<T> {
     pub start: Option<T>,
     pub end: Option<T>,
@@ -37,7 +38,7 @@ impl<T> Range<T> {
     }
 }
 
-#[derive(Debug, PartialEq, Eq, Clone, Serialize, Deserialize)]
+#[derive(Debug, PartialEq, Eq, Clone, Serialize, Deserialize, JsonSchema)]
 pub enum InterpolateItem<T> {
     String(String),
     Expr {
@@ -68,7 +69,7 @@ impl<T> InterpolateItem<T> {
     }
 }
 
-#[derive(Debug, PartialEq, Eq, Clone, Serialize, Deserialize)]
+#[derive(Debug, PartialEq, Eq, Clone, Serialize, Deserialize, JsonSchema)]
 pub struct SwitchCase<T> {
     pub condition: T,
     pub value: T,

prqlc/prqlc-parser/src/lexer/lr/token.rs

@@ -1,21 +1,23 @@
 use enum_as_inner::EnumAsInner;
+use schemars::JsonSchema;
 use serde::{Deserialize, Serialize};
 
-#[derive(Clone, PartialEq, Serialize, Deserialize, Eq)]
+#[derive(Clone, PartialEq, Serialize, Deserialize, Eq, JsonSchema)]
 pub struct Token {
     pub kind: TokenKind,
     pub span: std::ops::Range<usize>,
 }
 
-#[derive(Clone, PartialEq, Debug, Serialize, Deserialize)]
+#[derive(Clone, PartialEq, Debug, Serialize, Deserialize, JsonSchema)]
 pub enum TokenKind {
     NewLine,
 
     Ident(String),
     Keyword(String),
     #[cfg_attr(
         feature = "serde_yaml",
-        serde(with = "serde_yaml::with::singleton_map")
+        serde(with = "serde_yaml::with::singleton_map"),
+        schemars(with = "Literal")
     )]
     Literal(Literal),
     Param(String),
@@ -64,7 +66,9 @@ pub enum TokenKind {
     LineWrap(Vec<TokenKind>),
 }
 
-#[derive(Debug, EnumAsInner, PartialEq, Clone, Serialize, Deserialize, strum::AsRefStr)]
+#[derive(
+    Debug, EnumAsInner, PartialEq, Clone, Serialize, Deserialize, strum::AsRefStr, JsonSchema,
+)]
 pub enum Literal {
     Null,
     Integer(i64),
@@ -86,7 +90,7 @@ impl TokenKind {
     }
 }
 // Compound units, such as "2 days 3 hours" can be represented as `2days + 3hours`
-#[derive(Debug, Clone, Serialize, Deserialize, PartialEq, Eq)]
+#[derive(Debug, Clone, Serialize, Deserialize, PartialEq, Eq, JsonSchema)]
 pub struct ValueAndUnit {
     pub n: i64,       // Do any DBs use floats or decimals for this?
     pub unit: String, // Could be an enum IntervalType,

prqlc/prqlc-parser/src/parser/pr/expr.rs

@@ -1,6 +1,7 @@
 use std::collections::HashMap;
 
 use enum_as_inner::EnumAsInner;
+use schemars::JsonSchema;
 use serde::{Deserialize, Serialize};
 
 use crate::generic;
@@ -23,7 +24,7 @@ impl Expr {
 
 /// Expr is anything that has a value and thus a type.
 /// Most of these can contain other [Expr] themselves; literals should be [ExprKind::Literal].
-#[derive(Debug, Clone, PartialEq, Serialize, Deserialize)]
+#[derive(Debug, Clone, PartialEq, Serialize, Deserialize, JsonSchema)]
 pub struct Expr {
     #[serde(flatten)]
     pub kind: ExprKind,
@@ -35,7 +36,9 @@ pub struct Expr {
     pub alias: Option<String>,
 }
 
-#[derive(Debug, EnumAsInner, PartialEq, Clone, Serialize, Deserialize, strum::AsRefStr)]
+#[derive(
+    Debug, EnumAsInner, PartialEq, Clone, Serialize, Deserialize, strum::AsRefStr, JsonSchema,
+)]
 pub enum ExprKind {
     Ident(String),
 
@@ -47,7 +50,8 @@ pub enum ExprKind {
     },
     #[cfg_attr(
         feature = "serde_yaml",
-        serde(with = "serde_yaml::with::singleton_map")
+        serde(with = "serde_yaml::with::singleton_map"),
+        schemars(with = "Literal")
     )]
     Literal(Literal),
     Pipeline(Pipeline),
@@ -81,28 +85,28 @@ impl ExprKind {
     }
 }
 
-#[derive(Debug, EnumAsInner, PartialEq, Clone, Serialize, Deserialize)]
+#[derive(Debug, EnumAsInner, PartialEq, Clone, Serialize, Deserialize, JsonSchema)]
 pub enum IndirectionKind {
     Name(String),
     Position(i64),
     Star,
 }
 
-#[derive(Debug, PartialEq, Clone, Serialize, Deserialize)]
+#[derive(Debug, PartialEq, Clone, Serialize, Deserialize, JsonSchema)]
 pub struct BinaryExpr {
     pub left: Box<Expr>,
     pub op: BinOp,
     pub right: Box<Expr>,
 }
 
-#[derive(Debug, PartialEq, Clone, Serialize, Deserialize)]
+#[derive(Debug, PartialEq, Clone, Serialize, Deserialize, JsonSchema)]
 pub struct UnaryExpr {
     pub op: UnOp,
     pub expr: Box<Expr>,
 }
 
 /// Function call.
-#[derive(Debug, PartialEq, Clone, Serialize, Deserialize)]
+#[derive(Debug, PartialEq, Clone, Serialize, Deserialize, JsonSchema)]
 pub struct FuncCall {
     pub name: Box<Expr>,
     pub args: Vec<Expr>,
@@ -112,7 +116,7 @@ pub struct FuncCall {
 
 /// Function called with possibly missing positional arguments.
 /// May also contain environment that is needed to evaluate the body.
-#[derive(Debug, PartialEq, Clone, Serialize, Deserialize)]
+#[derive(Debug, PartialEq, Clone, Serialize, Deserialize, JsonSchema)]
 pub struct Func {
     /// Type requirement for the function body expression.
     pub return_ty: Option<Ty>,
@@ -130,7 +134,7 @@ pub struct Func {
     pub generic_type_params: Vec<GenericTypeParam>,
 }
 
-#[derive(Debug, PartialEq, Clone, Serialize, Deserialize)]
+#[derive(Debug, PartialEq, Clone, Serialize, Deserialize, JsonSchema)]
 pub struct FuncParam {
     pub name: String,
 
@@ -140,7 +144,7 @@ pub struct FuncParam {
     pub default_value: Option<Box<Expr>>,
 }
 
-#[derive(Debug, PartialEq, Clone, Serialize, Deserialize)]
+#[derive(Debug, PartialEq, Clone, Serialize, Deserialize, JsonSchema)]
 pub struct GenericTypeParam {
     /// Assigned name of this generic type argument.
     pub name: String,
@@ -149,7 +153,7 @@ pub struct GenericTypeParam {
 }
 
 /// A value and a series of functions that are to be applied to that value one after another.
-#[derive(Debug, PartialEq, Clone, Serialize, Deserialize)]
+#[derive(Debug, PartialEq, Clone, Serialize, Deserialize, JsonSchema)]
 pub struct Pipeline {
     pub exprs: Vec<Expr>,
 }

prqlc/prqlc-parser/src/parser/pr/ident.rs

@@ -1,10 +1,11 @@
 use std::fmt::Write;
 
+use schemars::JsonSchema;
 use serde::{ser::SerializeSeq, Deserialize, Deserializer, Serialize, Serializer};
 
 /// A name. Generally columns, tables, functions, variables.
 /// This is glorified way of writing a "vec with at least one element".
-#[derive(PartialEq, Eq, PartialOrd, Ord, Hash, Clone)]
+#[derive(PartialEq, Eq, PartialOrd, Ord, Hash, Clone, JsonSchema)]
 pub struct Ident {
     pub path: Vec<String>,
     pub name: String,

prqlc/prqlc-parser/src/parser/pr/ops.rs

@@ -1,3 +1,4 @@
+use schemars::JsonSchema;
 use serde::{Deserialize, Serialize};
 
 #[derive(
@@ -11,6 +12,7 @@ use serde::{Deserialize, Serialize};
     Deserialize,
     strum::Display,
     strum::EnumString,
+    JsonSchema,
 )]
 pub enum UnOp {
     #[strum(to_string = "-")]
@@ -34,6 +36,7 @@ pub enum UnOp {
     Deserialize,
     strum::Display,
     strum::EnumString,
+    JsonSchema,
 )]
 pub enum BinOp {
     #[strum(to_string = "*")]