jspg/src/lib.rs

use pgrx::*;

pg_module_magic!();

use boon::{CompileError, Compiler, ErrorKind, SchemaIndex, Schemas, ValidationError, Type, Types, ValidationOptions};
use lazy_static::lazy_static;
use serde_json::{json, Value, Number};
use std::borrow::Cow;
use std::collections::hash_map::Entry;
use std::{collections::{HashMap, HashSet}, sync::RwLock};

#[derive(Clone, Copy, Debug, PartialEq)]
enum SchemaType {
 Enum,
 Type,
 Family, // Added for generated hierarchy schemas
 PublicPunc,
 PrivatePunc,
}

struct Schema {
 index: SchemaIndex,
 t: SchemaType,
}

struct Cache {
 schemas: Schemas,
 map: HashMap<String, Schema>,
}

// Structure to hold error information without lifetimes
#[derive(Debug)]
struct Error {
 path: String,
 code: String,
 message: String,
 cause: Value, // Changed from String to Value to store JSON
}

lazy_static! {
 static ref SCHEMA_CACHE: RwLock<Cache> = RwLock::new(Cache {
  schemas: Schemas::new(),
  map: HashMap::new(),
 });
}

#[pg_extern(strict)]
fn cache_json_schemas(enums: JsonB, types: JsonB, puncs: JsonB) -> JsonB {
  let mut cache = SCHEMA_CACHE.write().unwrap();
  let enums_value: Value = enums.0;
  let types_value: Value = types.0;
  let puncs_value: Value = puncs.0;

  *cache = Cache {
    schemas: Schemas::new(),
    map: HashMap::new(),
  };

  let mut compiler = Compiler::new();
  compiler.enable_format_assertions();

  let mut errors = Vec::new();
  let mut schemas_to_compile = Vec::new();

  // Phase 1: Enums
  if let Some(enums_array) = enums_value.as_array() {
    for enum_row in enums_array {
      if let Some(schemas_raw) = enum_row.get("schemas") {
        if let Some(schemas_array) = schemas_raw.as_array() {
          for schema_def in schemas_array {
            if let Some(schema_id) = schema_def.get("$id").and_then(|v| v.as_str()) {
              schemas_to_compile.push((schema_id.to_string(), schema_def.clone(), SchemaType::Enum));
            }
          }
        }
      }
    }
  }

  // Phase 2: Types & Hierarchy Pre-processing
  let mut hierarchy_map: HashMap<String, HashSet<String>> = HashMap::new();
  if let Some(types_array) = types_value.as_array() {
    for type_row in types_array {
      // Process main schemas for the type
      if let Some(schemas_raw) = type_row.get("schemas") {
        if let Some(schemas_array) = schemas_raw.as_array() {
          for schema_def in schemas_array {
            if let Some(schema_id) = schema_def.get("$id").and_then(|v| v.as_str()) {
              schemas_to_compile.push((schema_id.to_string(), schema_def.clone(), SchemaType::Type));
            }
          }
        }
      }

      // Process hierarchy to build .family enums
      if let Some(type_name) = type_row.get("name").and_then(|v| v.as_str()) {
        if let Some(hierarchy_raw) = type_row.get("hierarchy") {
          if let Some(hierarchy_array) = hierarchy_raw.as_array() {
            for ancestor_val in hierarchy_array {
              if let Some(ancestor_name) = ancestor_val.as_str() {
                hierarchy_map
                  .entry(ancestor_name.to_string())
                  .or_default()
                  .insert(type_name.to_string());
              }
            }
          }
        }
      }
    }
  }

  // Generate and add the .family schemas
  for (base_type, descendant_types) in hierarchy_map {
    let family_schema_id = format!("{}.family", base_type);
    let enum_values: Vec<String> = descendant_types.into_iter().collect();
    let family_schema = json!({
      "$id": family_schema_id,
      "type": "string",
      "enum": enum_values
    });
    schemas_to_compile.push((family_schema_id, family_schema, SchemaType::Family));
  }

  // Phase 3: Puncs
  if let Some(puncs_array) = puncs_value.as_array() {
    for punc_row in puncs_array {
      if let Some(punc_obj) = punc_row.as_object() {
        if let Some(punc_name) = punc_obj.get("name").and_then(|v| v.as_str()) {
          let is_public = punc_obj.get("public").and_then(|v| v.as_bool()).unwrap_or(false);
          let punc_schema_type = if is_public { SchemaType::PublicPunc } else { SchemaType::PrivatePunc };
          if let Some(schemas_raw) = punc_obj.get("schemas") {
            if let Some(schemas_array) = schemas_raw.as_array() {
              for schema_def in schemas_array {
                if let Some(schema_id) = schema_def.get("$id").and_then(|v| v.as_str()) {
                  let request_schema_id = format!("{}.request", punc_name);
                  let response_schema_id = format!("{}.response", punc_name);
                  let schema_type_for_def = if schema_id == request_schema_id || schema_id == response_schema_id {
                    punc_schema_type
                  } else {
                    SchemaType::Type
                  };
                  schemas_to_compile.push((schema_id.to_string(), schema_def.clone(), schema_type_for_def));
                }
              }
            }
          }
        }
      }
    }
  }

  // Add all resources to compiler first
  for (id, value, schema_type) in &schemas_to_compile {
    add_schema_resource(&mut compiler, id, value.clone(), *schema_type, &mut errors);
  }

  if !errors.is_empty() {
    return JsonB(json!({ "errors": errors }));
  }

  // Compile all schemas
  compile_all_schemas(&mut compiler, &mut cache, &schemas_to_compile, &mut errors);

  if errors.is_empty() {
    JsonB(json!({ "response": "success" }))
  } else {
    JsonB(json!({ "errors": errors }))
  }
}

// Helper function to add a schema resource (without compiling)
fn add_schema_resource(
 compiler: &mut Compiler, 
 schema_id: &str, 
 schema_value: Value, 
 _schema_type: SchemaType,
 errors: &mut Vec<Value>
) {
 if let Err(e) = compiler.add_resource(schema_id, schema_value) {
  errors.push(json!({ 
   "code": "SCHEMA_RESOURCE_FAILED",
   "message": format!("Failed to add schema resource '{}'", schema_id),
   "details": { "schema": schema_id, "cause": format!("{}", e) }
  }));
 }
}

// Helper function to compile all added resources
fn compile_all_schemas(
  compiler: &mut Compiler,
  cache: &mut Cache,
  schemas_to_compile: &[(String, Value, SchemaType)],
  errors: &mut Vec<Value>,
) {
  for (id, value, schema_type) in schemas_to_compile {
    match compiler.compile(id, &mut cache.schemas) {
      Ok(index) => {
        cache.map.insert(id.clone(), Schema { index, t: *schema_type });
      }
      Err(e) => {
        match &e {
          CompileError::ValidationError { src, .. } => {
            let mut error_list = Vec::new();
            collect_errors(src, &mut error_list);
            let formatted_errors = format_errors(error_list, value, id);
            errors.extend(formatted_errors);
          }
          _ => {
            errors.push(json!({ 
              "code": "SCHEMA_COMPILATION_FAILED",
              "message": format!("Schema '{}' compilation failed", id),
              "details": { "schema": id, "cause": format!("{:?}", e) }
            }));
          }
        };
      }
    }
  }
}

#[pg_extern(strict, parallel_safe)]
fn validate_json_schema(schema_id: &str, instance: JsonB) -> JsonB {
  let cache = SCHEMA_CACHE.read().unwrap();
  match cache.map.get(schema_id) {
    None => JsonB(json!({ 
      "errors": [{
        "code": "SCHEMA_NOT_FOUND",
        "message": format!("Schema '{}' not found in cache", schema_id),
        "details": {
          "schema": schema_id,
          "cause": "Schema was not found in bulk cache - ensure cache_json_schemas was called"
        }
      }]
    })),
    Some(schema) => {
      let instance_value: Value = instance.0;
      let options = match schema.t {
        SchemaType::PublicPunc => Some(ValidationOptions { be_strict: true }),
        _ => None,
      };

      match cache.schemas.validate(&instance_value, schema.index, options) {
        Ok(_) => {
            JsonB(json!({ "response": "success" }))
        }
        Err(validation_error) => {
          let mut error_list = Vec::new();
          collect_errors(&validation_error, &mut error_list);
          let errors = format_errors(error_list, &instance_value, schema_id);
          if errors.is_empty() {
            JsonB(json!({ "response": "success" }))
          } else {
            JsonB(json!({ "errors": errors }))
          }
        }
      }
    }
  }
}

// Recursively collects validation errors
fn collect_errors(error: &ValidationError, errors_list: &mut Vec<Error>) {
    // Check if this is a structural error that we should skip
    let is_structural = matches!(
        &error.kind,
        ErrorKind::Group | ErrorKind::AllOf | ErrorKind::AnyOf | ErrorKind::Not | ErrorKind::OneOf(_)
    );

    if !error.causes.is_empty() || is_structural {
      for cause in &error.causes {
          collect_errors(cause, errors_list);
      }
      return
    }

    let base_path = error.instance_location.to_string();
    let errors_to_add = match &error.kind {
        ErrorKind::Type { got, want } => handle_type_error(&base_path, got, want),
        ErrorKind::Required { want } => handle_required_error(&base_path, want),
        ErrorKind::Dependency { prop, missing } => handle_dependency_error(&base_path, prop, missing, false),
        ErrorKind::DependentRequired { prop, missing } => handle_dependency_error(&base_path, prop, missing, true),
        ErrorKind::AdditionalProperties { got } => handle_additional_properties_error(&base_path, got),
        ErrorKind::Enum { want } => handle_enum_error(&base_path, want),
        ErrorKind::Const { want } => handle_const_error(&base_path, want),
        ErrorKind::MinLength { got, want } => handle_min_length_error(&base_path, *got, *want),
        ErrorKind::MaxLength { got, want } => handle_max_length_error(&base_path, *got, *want),
        ErrorKind::Pattern { got, want } => handle_pattern_error(&base_path, got, want),
        ErrorKind::Minimum { got, want } => handle_minimum_error(&base_path, got, want),
        ErrorKind::Maximum { got, want } => handle_maximum_error(&base_path, got, want),
        ErrorKind::ExclusiveMinimum { got, want } => handle_exclusive_minimum_error(&base_path, got, want),
        ErrorKind::ExclusiveMaximum { got, want } => handle_exclusive_maximum_error(&base_path, got, want),
        ErrorKind::MultipleOf { got, want } => handle_multiple_of_error(&base_path, got, want),
        ErrorKind::MinItems { got, want } => handle_min_items_error(&base_path, *got, *want),
        ErrorKind::MaxItems { got, want } => handle_max_items_error(&base_path, *got, *want),
        ErrorKind::UniqueItems { got } => handle_unique_items_error(&base_path, got),
        ErrorKind::MinProperties { got, want } => handle_min_properties_error(&base_path, *got, *want),
        ErrorKind::MaxProperties { got, want } => handle_max_properties_error(&base_path, *got, *want),
        ErrorKind::AdditionalItems { got } => handle_additional_items_error(&base_path, *got),
        ErrorKind::Format { want, got, err } => handle_format_error(&base_path, want, got, err),
        ErrorKind::PropertyName { prop } => handle_property_name_error(&base_path, prop),
        ErrorKind::Contains => handle_contains_error(&base_path),
        ErrorKind::MinContains { got, want } => handle_min_contains_error(&base_path, got, *want),
        ErrorKind::MaxContains { got, want } => handle_max_contains_error(&base_path, got, *want),
        ErrorKind::ContentEncoding { want, err } => handle_content_encoding_error(&base_path, want, err),
        ErrorKind::ContentMediaType { want, err, .. } => handle_content_media_type_error(&base_path, want, err),
        ErrorKind::FalseSchema => handle_false_schema_error(&base_path),
        ErrorKind::Not => handle_not_error(&base_path),
        ErrorKind::RefCycle { url, kw_loc1, kw_loc2 } => handle_ref_cycle_error(&base_path, url, kw_loc1, kw_loc2),
        ErrorKind::Reference { kw, url } => handle_reference_error(&base_path, kw, url),
        ErrorKind::Schema { url } => handle_schema_error(&base_path, url),
        ErrorKind::ContentSchema => handle_content_schema_error(&base_path),
        ErrorKind::Group => handle_group_error(&base_path),
        ErrorKind::AllOf => handle_all_of_error(&base_path),
        ErrorKind::AnyOf => handle_any_of_error(&base_path),
        ErrorKind::OneOf(matched) => handle_one_of_error(&base_path, matched),
    };

    errors_list.extend(errors_to_add);
}

// Handler functions for each error kind
fn handle_type_error(base_path: &str, got: &Type, want: &Types) -> Vec<Error> {
 vec![Error {
  path: base_path.to_string(),
  code: "TYPE_MISMATCH".to_string(),
  message: format!("Expected {} but got {}", 
   want.iter().map(|t| t.to_string()).collect::<Vec<_>>().join(" or "), 
   got
  ),
  cause: json!({ 
   "got": got.to_string(), 
   "want": want.iter().map(|t| t.to_string()).collect::<Vec<_>>() 
  }),
 }]
}

fn handle_required_error(base_path: &str, want: &[&str]) -> Vec<Error> {
 // Create a separate error for each missing required field
 want.iter().map(|missing_field| {
  let field_path = if base_path.is_empty() {
   format!("/{}", missing_field)
  } else {
   format!("{}/{}", base_path, missing_field)
  };
  
  Error {
   path: field_path,
   code: "REQUIRED_FIELD_MISSING".to_string(),
   message: format!("Required field '{}' is missing", missing_field),
   cause: json!({ "want": [missing_field] }),
  }
 }).collect()
}

fn handle_dependency_error(base_path: &str, prop: &str, missing: &[&str], is_dependent_required: bool) -> Vec<Error> {
 // Create a separate error for each missing field
 missing.iter().map(|missing_field| {
  let field_path = if base_path.is_empty() {
   format!("/{}", missing_field)
  } else {
   format!("{}/{}", base_path, missing_field)
  };
  
  let (code, message) = if is_dependent_required {
   (
    "DEPENDENT_REQUIRED_MISSING".to_string(),
    format!("Field '{}' is required when '{}' is present", missing_field, prop),
   )
  } else {
   (
    "DEPENDENCY_FAILED".to_string(),
    format!("Field '{}' is required when '{}' is present", missing_field, prop),
   )
  };
  
  Error {
   path: field_path,
   code,
   message,
   cause: json!({ "prop": prop, "missing": [missing_field] }),
  }
 }).collect()
}

fn handle_additional_properties_error(base_path: &str, got: &[Cow<str>]) -> Vec<Error> {
    let mut errors = Vec::new();
    for extra_prop in got {
        let field_path = if base_path.is_empty() {
            format!("/{}", extra_prop)
        } else {
            format!("{}/{}", base_path, extra_prop)
        };
        errors.push(Error {
            path: field_path,
            code: "ADDITIONAL_PROPERTIES_NOT_ALLOWED".to_string(),
            message: format!("Property '{}' is not allowed", extra_prop),
            cause: json!({ "got": [extra_prop.to_string()] }),
        });
    }
    errors
}

fn handle_enum_error(base_path: &str, want: &[Value]) -> Vec<Error> {
 let message = if want.len() == 1 {
  format!("Value must be {}", serde_json::to_string(&want[0]).unwrap_or_else(|_| "unknown".to_string()))
 } else {
  format!("Value must be one of: {}", 
   want.iter()
    .map(|v| serde_json::to_string(v).unwrap_or_else(|_| "unknown".to_string()))
    .collect::<Vec<_>>()
    .join(", ")
  )
 };
 
 vec![Error {
  path: base_path.to_string(),
  code: "ENUM_VIOLATED".to_string(),
  message,
  cause: json!({ "want": want }),
 }]
}

fn handle_const_error(base_path: &str, want: &Value) -> Vec<Error> {
 vec![Error {
  path: base_path.to_string(),
  code: "CONST_VIOLATED".to_string(),
  message: format!("Value must be exactly {}", serde_json::to_string(want).unwrap_or_else(|_| "unknown".to_string())),
  cause: json!({ "want": want }),
 }]
}

fn handle_min_length_error(base_path: &str, got: usize, want: usize) -> Vec<Error> {
 vec![Error {
  path: base_path.to_string(),
  code: "MIN_LENGTH_VIOLATED".to_string(),
  message: format!("String length must be at least {} characters, but got {}", want, got),
  cause: json!({ "got": got, "want": want }),
 }]
}

fn handle_max_length_error(base_path: &str, got: usize, want: usize) -> Vec<Error> {
 vec![Error {
  path: base_path.to_string(),
  code: "MAX_LENGTH_VIOLATED".to_string(),
  message: format!("String length must be at most {} characters, but got {}", want, got),
  cause: json!({ "got": got, "want": want }),
 }]
}

fn handle_pattern_error(base_path: &str, got: &Cow<str>, want: &str) -> Vec<Error> {
 let display_value = if got.len() > 50 { 
  format!("{}...", &got[..50]) 
 } else { 
  got.to_string() 
 };
 
 vec![Error {
  path: base_path.to_string(),
  code: "PATTERN_VIOLATED".to_string(),
  message: format!("Value '{}' does not match pattern '{}'", display_value, want),
  cause: json!({ "got": got.to_string(), "want": want }),
 }]
}

fn handle_minimum_error(base_path: &str, got: &Cow<Number>, want: &Number) -> Vec<Error> {
 vec![Error {
  path: base_path.to_string(),
  code: "MINIMUM_VIOLATED".to_string(),
  message: format!("Value must be at least {}, but got {}", want, got),
  cause: json!({ "got": got, "want": want }),
 }]
}

fn handle_maximum_error(base_path: &str, got: &Cow<Number>, want: &Number) -> Vec<Error> {
 vec![Error {
  path: base_path.to_string(),
  code: "MAXIMUM_VIOLATED".to_string(),
  message: format!("Value must be at most {}, but got {}", want, got),
  cause: json!({ "got": got, "want": want }),
 }]
}

fn handle_exclusive_minimum_error(base_path: &str, got: &Cow<Number>, want: &Number) -> Vec<Error> {
 vec![Error {
  path: base_path.to_string(),
  code: "EXCLUSIVE_MINIMUM_VIOLATED".to_string(),
  message: format!("Value must be greater than {}, but got {}", want, got),
  cause: json!({ "got": got, "want": want }),
 }]
}

fn handle_exclusive_maximum_error(base_path: &str, got: &Cow<Number>, want: &Number) -> Vec<Error> {
 vec![Error {
  path: base_path.to_string(),
  code: "EXCLUSIVE_MAXIMUM_VIOLATED".to_string(),
  message: format!("Value must be less than {}, but got {}", want, got),
  cause: json!({ "got": got, "want": want }),
 }]
}

fn handle_multiple_of_error(base_path: &str, got: &Cow<Number>, want: &Number) -> Vec<Error> {
 vec![Error {
  path: base_path.to_string(),
  code: "MULTIPLE_OF_VIOLATED".to_string(),
  message: format!("{} is not a multiple of {}", got, want),
  cause: json!({ "got": got, "want": want }),
 }]
}

fn handle_min_items_error(base_path: &str, got: usize, want: usize) -> Vec<Error> {
 vec![Error {
  path: base_path.to_string(),
  code: "MIN_ITEMS_VIOLATED".to_string(),
  message: format!("Array must have at least {} items, but has {}", want, got),
  cause: json!({ "got": got, "want": want }),
 }]
}

fn handle_max_items_error(base_path: &str, got: usize, want: usize) -> Vec<Error> {
 vec![Error {
  path: base_path.to_string(),
  code: "MAX_ITEMS_VIOLATED".to_string(),
  message: format!("Array must have at most {} items, but has {}", want, got),
  cause: json!({ "got": got, "want": want }),
 }]
}

fn handle_unique_items_error(base_path: &str, got: &[usize; 2]) -> Vec<Error> {
 vec![Error {
  path: base_path.to_string(),
  code: "UNIQUE_ITEMS_VIOLATED".to_string(),
  message: format!("Array items at positions {} and {} are duplicates", got[0], got[1]),
  cause: json!({ "got": got }),
 }]
}

fn handle_min_properties_error(base_path: &str, got: usize, want: usize) -> Vec<Error> {
 vec![Error {
  path: base_path.to_string(),
  code: "MIN_PROPERTIES_VIOLATED".to_string(),
  message: format!("Object must have at least {} properties, but has {}", want, got),
  cause: json!({ "got": got, "want": want }),
 }]
}

fn handle_max_properties_error(base_path: &str, got: usize, want: usize) -> Vec<Error> {
 vec![Error {
  path: base_path.to_string(),
  code: "MAX_PROPERTIES_VIOLATED".to_string(),
  message: format!("Object must have at most {} properties, but has {}", want, got),
  cause: json!({ "got": got, "want": want }),
 }]
}

fn handle_additional_items_error(base_path: &str, got: usize) -> Vec<Error> {
 vec![Error {
  path: base_path.to_string(),
  code: "ADDITIONAL_ITEMS_NOT_ALLOWED".to_string(),
  message: format!("Last {} array items are not allowed", got),
  cause: json!({ "got": got }),
 }]
}

fn handle_format_error(base_path: &str, want: &str, got: &Cow<Value>, err: &Box<dyn std::error::Error>) -> Vec<Error> {
 // If the value is an empty string, skip format validation.
 if let Value::String(s) = got.as_ref() {
  if s.is_empty() {
   return vec![];
  }
 }

 vec![Error {
  path: base_path.to_string(),
  code: "FORMAT_INVALID".to_string(),
  message: format!("Value {} is not a valid {} format", 
   serde_json::to_string(got.as_ref()).unwrap_or_else(|_| "unknown".to_string()),
   want
  ),
  cause: json!({ "got": got, "want": want, "err": err.to_string() }),
 }]
}

fn handle_property_name_error(base_path: &str, prop: &str) -> Vec<Error> {
 vec![Error {
  path: base_path.to_string(),
  code: "INVALID_PROPERTY_NAME".to_string(),
  message: format!("Property name '{}' is invalid", prop),
  cause: json!({ "prop": prop }),
 }]
}

fn handle_contains_error(base_path: &str) -> Vec<Error> {
 vec![Error {
  path: base_path.to_string(),
  code: "CONTAINS_FAILED".to_string(),
  message: "No array items match the required schema".to_string(),
  cause: json!({}),
 }]
}

fn handle_min_contains_error(base_path: &str, got: &[usize], want: usize) -> Vec<Error> {
 let message = if got.is_empty() {
  format!("At least {} array items must match the schema, but none do", want)
 } else {
  format!("At least {} array items must match the schema, but only {} do (at positions {})", 
   want, 
   got.len(), 
   got.iter().map(|i| i.to_string()).collect::<Vec<_>>().join(", ")
  )
 };
 
 vec![Error {
  path: base_path.to_string(),
  code: "MIN_CONTAINS_VIOLATED".to_string(),
  message,
  cause: json!({ "got": got, "want": want }),
 }]
}

fn handle_max_contains_error(base_path: &str, got: &[usize], want: usize) -> Vec<Error> {
 vec![Error {
  path: base_path.to_string(),
  code: "MAX_CONTAINS_VIOLATED".to_string(),
  message: format!("At most {} array items can match the schema, but {} do (at positions {})", 
   want, 
   got.len(), 
   got.iter().map(|i| i.to_string()).collect::<Vec<_>>().join(", ")
  ),
  cause: json!({ "got": got, "want": want }),
 }]
}

fn handle_content_encoding_error(base_path: &str, want: &str, err: &Box<dyn std::error::Error>) -> Vec<Error> {
 vec![Error {
  path: base_path.to_string(),
  code: "CONTENT_ENCODING_INVALID".to_string(),
  message: format!("Content is not valid {} encoding: {}", want, err),
  cause: json!({ "want": want, "err": err.to_string() }),
 }]
}

fn handle_content_media_type_error(base_path: &str, want: &str, err: &Box<dyn std::error::Error>) -> Vec<Error> {
 vec![Error {
  path: base_path.to_string(),
  code: "CONTENT_MEDIA_TYPE_INVALID".to_string(),
  message: format!("Content is not valid {} media type: {}", want, err),
  cause: json!({ "want": want, "err": err.to_string() }),
 }]
}

fn handle_false_schema_error(base_path: &str) -> Vec<Error> {
 vec![Error {
  path: base_path.to_string(),
  code: "FALSE_SCHEMA".to_string(),
  message: "This schema always fails validation".to_string(),
  cause: json!({}),
 }]
}

fn handle_not_error(base_path: &str) -> Vec<Error> {
 vec![Error {
  path: base_path.to_string(),
  code: "NOT_VIOLATED".to_string(),
  message: "Value matches a schema that it should not match".to_string(),
  cause: json!({}),
 }]
}

fn handle_ref_cycle_error(base_path: &str, url: &str, kw_loc1: &str, kw_loc2: &str) -> Vec<Error> {
 vec![Error {
  path: base_path.to_string(),
  code: "REFERENCE_CYCLE".to_string(),
  message: format!("Reference cycle detected: both '{}' and '{}' resolve to '{}'", kw_loc1, kw_loc2, url),
  cause: json!({ "url": url, "kw_loc1": kw_loc1, "kw_loc2": kw_loc2 }),
 }]
}

fn handle_reference_error(base_path: &str, kw: &str, url: &str) -> Vec<Error> {
 vec![Error {
  path: base_path.to_string(),
  code: "REFERENCE_FAILED".to_string(),
  message: format!("{} reference to '{}' failed validation", kw, url),
  cause: json!({ "kw": kw, "url": url }),
 }]
}

fn handle_schema_error(base_path: &str, url: &str) -> Vec<Error> {
 vec![Error {
  path: base_path.to_string(),
  code: "SCHEMA_FAILED".to_string(),
  message: format!("Schema '{}' validation failed", url),
  cause: json!({ "url": url }),
 }]
}

fn handle_content_schema_error(base_path: &str) -> Vec<Error> {
 vec![Error {
  path: base_path.to_string(),
  code: "CONTENT_SCHEMA_FAILED".to_string(),
  message: "Content schema validation failed".to_string(),
  cause: json!({}),
 }]
}

fn handle_group_error(base_path: &str) -> Vec<Error> {
 vec![Error {
  path: base_path.to_string(),
  code: "VALIDATION_FAILED".to_string(),
  message: "Validation failed".to_string(),
  cause: json!({}),
 }]
}

fn handle_all_of_error(base_path: &str) -> Vec<Error> {
 vec![Error {
  path: base_path.to_string(),
  code: "ALL_OF_VIOLATED".to_string(),
  message: "Value does not match all required schemas".to_string(),
  cause: json!({}),
 }]
}

fn handle_any_of_error(base_path: &str) -> Vec<Error> {
 vec![Error {
  path: base_path.to_string(),
  code: "ANY_OF_VIOLATED".to_string(),
  message: "Value does not match any of the allowed schemas".to_string(),
  cause: json!({}),
 }]
}

fn handle_one_of_error(base_path: &str, matched: &Option<(usize, usize)>) -> Vec<Error> {
 let (message, cause) = match matched {
  None => (
   "Value must match exactly one schema, but matches none".to_string(),
   json!({ "matched_indices": null })
  ),
  Some((i, j)) => (
   format!("Value must match exactly one schema, but matches schemas at positions {} and {}", i, j),
   json!({ "matched_indices": [i, j] })
  ),
 };
 
 vec![Error {
  path: base_path.to_string(),
  code: "ONE_OF_VIOLATED".to_string(),
  message,
  cause,
 }]
}

// Formats errors according to DropError structure
fn format_errors(errors: Vec<Error>, instance: &Value, schema_id: &str) -> Vec<Value> {
    let mut unique_errors: HashMap<String, Value> = HashMap::new();
    for error in errors {
        let error_path = error.path.clone();
        if let Entry::Vacant(entry) = unique_errors.entry(error_path.clone()) {
            let failing_value = extract_value_at_path(instance, &error.path);
            entry.insert(json!({
                "code": error.code,
                "message": error.message,
                "details": {
                    "path": error.path,
                    "context": failing_value,
                    "cause": error.cause,
                    "schema": schema_id
                }
            }));
        }
    }
    
    unique_errors.into_values().collect::<Vec<Value>>()
}

// Helper function to extract value at a JSON pointer path
fn extract_value_at_path(instance: &Value, path: &str) -> Value {
 let parts: Vec<&str> = path.split('/').filter(|s| !s.is_empty()).collect();
 let mut current = instance;

 for part in parts {
  match current {
   Value::Object(map) => {
    if let Some(value) = map.get(part) {
     current = value;
    } else {
     return Value::Null;
    }
   }
   Value::Array(arr) => {
    if let Ok(index) = part.parse::<usize>() {
     if let Some(value) = arr.get(index) {
      current = value;
     } else {
      return Value::Null;
     }
    } else {
     return Value::Null;
    }
   }
   _ => return Value::Null,
  }
 }

 current.clone()
}

#[pg_extern(strict, parallel_safe)]
fn json_schema_cached(schema_id: &str) -> bool {
 let cache = SCHEMA_CACHE.read().unwrap();
 cache.map.contains_key(schema_id)
}

#[pg_extern(strict)]
fn clear_json_schemas() -> JsonB {
 let mut cache = SCHEMA_CACHE.write().unwrap();
 *cache = Cache {
  schemas: Schemas::new(),
  map: HashMap::new(),
 };
 JsonB(json!({ "response": "success" }))
}

#[pg_extern(strict, parallel_safe)]
fn show_json_schemas() -> JsonB {
 let cache = SCHEMA_CACHE.read().unwrap();
 let ids: Vec<String> = cache.map.keys().cloned().collect();
 JsonB(json!({ "response": ids }))
}

/// This module is required by `cargo pgrx test` invocations.
/// It must be visible at the root of your extension crate.
#[cfg(test)]
pub mod pg_test {
 pub fn setup(_options: Vec<&str>) {
  // perform one-off initialization when the pg_test framework starts
 }

 #[must_use]
 pub fn postgresql_conf_options() -> Vec<&'static str> {
  // return any postgresql.conf settings that are required for your tests
  vec![]
 }
}

#[cfg(any(test, feature = "pg_test"))]
mod helpers {
 include!("helpers.rs");
}

#[cfg(any(test, feature = "pg_test"))]
mod schemas {
 include!("schemas.rs");
}

#[cfg(any(test, feature = "pg_test"))]
#[pg_schema]
mod tests {
 include!("tests.rs");
}