util.jsonschema: Fix validation to not assume presence of "type" field

MattJ reported a curious issue where validation did not work as
expected. Primarily that the "type" field was expected to be mandatory,
and thus leaving it out would result in no checks being performed.
This was likely caused by misreading during initial development.

Spent some time testing against
https://github.com/json-schema-org/JSON-Schema-Test-Suite.git and
discovered a multitude of issues, far too many to bother splitting into
separate commits.

More than half of them fail. Many because of features not implemented,
which have been marked NYI. For example, some require deep comparisons
e.g. when objects or arrays are present in enums fields.

Some because of quirks with how Lua differs from JavaScript, e.g. no
distinct array or object types. Tests involving fractional floating
point numbers. We're definitely not going to follow references to remote
resources. Or deal with UTF-16 sillyness. One test asserted that 1.0 is
an integer, where Lua 5.3+ will disagree.

spec/util_jsonschema_spec.lua

@@ -0,0 +1,102 @@
+local js = require "util.jsonschema";
+local json = require "util.json";
+local lfs = require "lfs";
+
+-- https://github.com/json-schema-org/JSON-Schema-Test-Suite.git 2.0.0-550-g88d6948
+local test_suite_dir = "spec/JSON-Schema-Test-Suite/tests/draft2020-12"
+if lfs.attributes(test_suite_dir, "mode") ~= "directory" then return end
+
+-- Tests to skip and short reason why (NYI = not yet implemented)
+local skip = {
+	["ref.json:0:3"] = "NYI additionalProperties";
+	["ref.json:3:2"] = "FIXME investigate, util.jsonpath issue?",
+	["ref.json:6:1"] = "NYI",
+	["required.json:0:2"] = "distinguishing objects from arrays",
+	["additionalProperties.json:0:2"] = "distinguishing objects from arrays",
+	["additionalProperties.json:0:5"] = "NYI",
+	["additionalProperties.json:1:0"] = "NYI",
+	["anchor.json"] = "$anchor NYI",
+	["const.json:1"] = "deepcompare",
+	["const.json:13:2"] = "IEEE 754 equality",
+	["const.json:2"] = "deepcompare",
+	["const.json:8"] = "deepcompare",
+	["const.json:9"] = "deepcompare",
+	["contains.json:0:5"] = "distinguishing objects from arrays",
+	["defs.json"] = "need built-in meta-schema",
+	["dependentRequired.json"] = "NYI",
+	["dependentSchemas.json"] = "NYI",
+	["dynamicRef.json"] = "NYI",
+	["enum.json:1:3"] = "deepcompare",
+	["id.json"] = "NYI",
+	["maxContains.json"] = "NYI",
+	["maxLength.json:0:4"] = "UTF-16",
+	["maxProperties.json"] = "NYI",
+	["minContains.json"] = "NYI",
+	["minLength.json:0:4"] = "UTF-16",
+	["minProperties.json"] = "NYI",
+	["multipleOf.json:1"] = "multiples of IEEE 754 fractions",
+	["multipleOf.json:2"] = "multiples of IEEE 754 fractions",
+	["pattern.json"] = "NYI",
+	["patternProperties.json"] = "NYI",
+	["properties.json:1:2"] = "NYI",
+	["properties.json:1:3"] = "NYI",
+	["ref.json:14"] = "NYI",
+	["ref.json:15"] = "NYI",
+	["ref.json:16"] = "NYI",
+	["ref.json:17"] = "NYI",
+	["ref.json:18"] = "NYI",
+	["ref.json:13"] = "NYI",
+	["ref.json:19"] = "NYI",
+	["ref.json:11"] = "NYI",
+	["ref.json:12:1"] = "FIXME",
+	["refRemote.json"] = "DEFINITELY NYI",
+	["type.json:3:4"] = "distinguishing objects from arrays",
+	["type.json:3:6"] = "null is weird",
+	["type.json:4:3"] = "distinguishing objects from arrays",
+	["type.json:4:6"] = "null is weird",
+	["type.json:9:4"] = "null is weird",
+	["type.json:9:6"] = "null is weird",
+	["unevaluatedItems.json"] = "NYI",
+	["unevaluatedProperties.json"] = "NYI",
+	["uniqueItems.json:0:11"] = "deepcompare",
+	["uniqueItems.json:0:13"] = "deepcompare",
+	["uniqueItems.json:0:14"] = "deepcompare",
+	["uniqueItems.json:0:22"] = "deepcompare",
+	["uniqueItems.json:0:24"] = "deepcompare",
+	["uniqueItems.json:0:9"] = "deepcompare",
+	["unknownKeyword.json"] = "NYI",
+	["vocabulary.json"] = "NYI",
+};
+
+local function label(s, i)
+	return string.format("%s:%d", s, i-1);
+end
+
+describe("util.jsonschema.validate", function()
+	for test_case_file in lfs.dir(test_suite_dir) do
+		-- print(skip[test_case_file] and "do  " or "skip", test_case_file)
+		if test_case_file:sub(-5) == ".json" and not skip[test_case_file] then
+			describe(test_case_file, function()
+				local test_cases;
+				setup(function()
+					local f = assert(io.open(test_suite_dir .. "/" .. test_case_file));
+					local rawdata = assert(f:read("*a"), "failed to read " .. test_case_file)
+					test_cases = assert(json.decode(rawdata), "failed to parse " .. test_case_file)
+				end)
+				describe("tests", function()
+					for i, schema_test in ipairs(test_cases) do
+						local generic_label = label(test_case_file, i);
+						describe(schema_test.description or generic_label, function()
+							for j, test in ipairs(schema_test.tests) do
+								local specific_label = label(generic_label, j);
+								((skip[generic_label] or skip[specific_label]) and pending or it)(test.description, function()
+									assert.equal(test.valid, js.validate(schema_test.schema, test.data), specific_label .. " " .. test.description);
+								end)
+							end
+						end)
+					end
+				end)
+			end)
+		end
+	end
+end);

teal-src/util/jsonschema.tl

@@ -16,13 +16,13 @@ local pointer = require "util.jsonpointer"
 local type json_type_name = json.json_type_name
 
 -- json_type_name here is non-standard
-local type schema_t = boolean | json_type_name | json_schema_object
+local type schema_t = boolean | json_schema_object
 
 local record json_schema_object
 	type json_type_name = json.json_type_name
 	type schema_object = json_schema_object
 
-	type : json_type_name
+	type : json_type_name | { json_type_name }
 	enum : { any }
 	const : any
 
@@ -47,7 +47,7 @@ local record json_schema_object
 	-- strings
 	maxLength : integer
 	minLength : integer
-	pattern : string
+	pattern : string -- NYI
 	format : string
 
 	-- arrays
@@ -57,17 +57,17 @@ local record json_schema_object
 	maxItems : integer
 	minItems : integer
 	uniqueItems : boolean
-	maxContains : integer
+	maxContains : integer -- NYI
-	minContains : integer
+	minContains : integer -- NYI
 
 	-- objects
 	properties : { string : schema_t }
-	maxProperties : integer
+	maxProperties : integer -- NYI
-	minProperties : integer
+	minProperties : integer -- NYI
 	required : { string }
 	dependentRequired : { string : { string } }
 	additionalProperties: schema_t
-	patternProperties: schema_t
+	patternProperties: schema_t -- NYI
 	propertyNames : schema_t
 
 	-- xml
@@ -99,10 +99,10 @@ end
 
 -- TODO validator function per schema property
 
-local type_validators : { json_type_name : function (schema_t, any, json_schema_object) : boolean } = {}
+local function simple_validate(schema : json_type_name | { json_type_name }, data : any) : boolean
-
+	if schema == nil then
-local function simple_validate(schema : json_type_name, data : any) : boolean
+		return true
-	if schema == "object" and data is table then
+	elseif schema == "object" and data is table then
 		return type(data) == "table" and (next(data)==nil or type((next(data, nil))) == "string")
 	elseif schema == "array" and data is table then
 		return type(data) == "table" and (next(data)==nil or type((next(data, nil))) == "number")
@@ -110,12 +110,83 @@ local function simple_validate(schema : json_type_name, data : any) : boolean
 		return math.type(data) == schema
 	elseif schema == "null" then
 		return data == null
+	elseif schema is { json_type_name } then
+		for _, one in ipairs(schema as { json_type_name }) do
+			if simple_validate(one, data) then
+				return true
+			end
+		end
+		return false
 	else
 		return type(data) == schema
 	end
 end
 
-type_validators.string = function (schema : json_schema_object, data : any) : boolean
+local complex_validate : function ( json_schema_object, any, json_schema_object ) : boolean
+
+local function validate (schema : schema_t, data : any, root : json_schema_object) : boolean
+	if schema is boolean then
+		return schema
+	else
+		return complex_validate(schema, data, root)
+	end
+end
+
+function complex_validate (schema : json_schema_object, data : any, root : json_schema_object) : boolean
+
+	if root == nil then
+		root = schema
+	end
+
+	if schema["$ref"] and schema["$ref"]:sub(1,1) == "#" then
+		local referenced = pointer.resolve(root as table, schema["$ref"]:sub(2)) as schema_t
+		if referenced ~= nil and referenced ~= root and referenced ~= schema then
+			if not validate(referenced, data, root) then
+				return false;
+			end
+		end
+	end
+
+	if not simple_validate(schema.type, data) then
+		return false;
+	end
+
+	if schema.type == "object" then
+		if data is table then
+			-- just check that there the keys are all strings
+			for k in pairs(data) do
+				if not k is string then
+					return false
+				end
+			end
+		end
+	end
+
+	if schema.type == "array" then
+		if data is table then
+			-- just check that there the keys are all numbers
+			for i in pairs(data) do
+				if not i is integer then
+					return false
+				end
+			end
+		end
+	end
+
+	if schema["enum"] ~= nil then
+		local match = false
+		for _, v in ipairs(schema["enum"]) do
+			if v == data then
+				-- FIXME supposed to do deep-compare
+				match = true
+				break
+			end
+		end
+		if not match then
+			return false
+		end
+	end
+
 	-- XXX this is measured in byte, while JSON measures in ... bork
 	-- TODO use utf8.len?
 	if data is string then
@@ -125,13 +196,10 @@ type_validators.string = function (schema : json_schema_object, data : any) : bo
 		if schema.minLength and #data < schema.minLength then
 			return false
 		end
-		return true
 	end
-	return false
-end
 
-type_validators.number = function (schema : json_schema_object, data : number) : boolean
+	if data is number then
-	if schema.multipleOf and data % schema.multipleOf ~= 0 then
+		if schema.multipleOf and (data == 0 or data % schema.multipleOf ~= 0) then
 			return false
 		end
 
@@ -150,28 +218,6 @@ type_validators.number = function (schema : json_schema_object, data : number) :
 		if schema.exclusiveMinimum and not ( data > schema.exclusiveMinimum ) then
 			return false
 		end
-
-	return true
-end
-
-type_validators.integer = type_validators.number
-
-local function validate(schema : schema_t, data : any, root : json_schema_object) : boolean
-	if schema is boolean then
-		return schema
-	end
-	if schema is json_type_name then
-		return simple_validate(schema, data)
-	end
-	if schema is json_schema_object then
-		if root == nil then
-			root = schema
-		end
-		if schema["$ref"] and schema["$ref"]:sub(1,1) == "#" then
-			local referenced = pointer.resolve(root as table, schema["$ref"]:sub(2)) as schema_t
-			if referenced ~= nil then
-				return validate(referenced, data, root);
-			end
 	end
 
 	if schema.allOf then
@@ -180,7 +226,6 @@ local function validate(schema : schema_t, data : any, root : json_schema_object
 				return false
 			end
 		end
-			return true
 	end
 
 	if schema.oneOf then
@@ -190,17 +235,23 @@ local function validate(schema : schema_t, data : any, root : json_schema_object
 				valid = valid + 1
 			end
 		end
-			return valid == 1
+		if valid ~= 1 then
+			return false
+		end
 	end
 
 	if schema.anyOf then
+		local match = false
 		for _, sub in ipairs(schema.anyOf) do
 			if validate(sub, data, root) then
-					return true
+				match = true
+				break
 			end
 		end
+		if not match then
 			return false
 		end
+	end
 
 	if schema["not"] then
 		if validate(schema["not"], data, root) then
@@ -208,7 +259,7 @@ local function validate(schema : schema_t, data : any, root : json_schema_object
 		end
 	end
 
-		if schema["if"] then
+	if schema["if"] ~= nil then
 		if validate(schema["if"], data, root) then
 			if schema["then"] then
 				return validate(schema["then"], data, root)
@@ -224,30 +275,6 @@ local function validate(schema : schema_t, data : any, root : json_schema_object
 		return false
 	end
 
-		if schema["enum"] ~= nil then
-			for _, v in ipairs(schema["enum"]) do
-				if v == data then
-					return true
-				end
-			end
-			return false
-		end
-
-		if schema.type then
-			if not simple_validate(schema.type, data) then
-				return false
-			end
-
-			local validator = type_validators[schema.type]
-			if validator then
-				return validator(schema, data, root)
-			end
-		end
-		return true
-	end
-end
-
-type_validators.table = function (schema : json_schema_object, data : any, root : json_schema_object) : boolean
 	if data is table then
 
 		if schema.maxItems and #data > schema.maxItems then
@@ -266,27 +293,31 @@ type_validators.table = function (schema : json_schema_object, data : any, root
 			end
 		end
 
+		if schema.propertyNames ~= nil then
+			for k in pairs(data) do
+				if not validate(schema.propertyNames, k, root) then
+					return false
+				end
+			end
+		end
+
 		if schema.properties then
-			local additional : schema_t = schema.additionalProperties or true
+			for k, sub in pairs(schema.properties) do
+				if data[k] ~= nil and not validate(sub, data[k], root) then
+					return false
+				end
+			end
+		end
+
+		if schema.additionalProperties ~= nil then
 			for k, v in pairs(data) do
-				if schema.propertyNames and not validate(schema.propertyNames, k, root) then
+				if schema.properties == nil or schema.properties[k as string] == nil then
-					return false
-				end
-				local s = schema.properties[k as string] or additional
-				if not validate(s, v, root) then
-					return false
-				end
-			end
-		elseif schema.additionalProperties then
-			for k, v in pairs(data) do
-				if schema.propertyNames and not validate(schema.propertyNames, k, root) then
-					return false
-				end
 					if not validate(schema.additionalProperties, v, root) then
 						return false
 					end
 				end
 			end
+		end
 
 		if schema.uniqueItems then
 			-- only works for scalars, would need to deep-compare for objects/arrays/tables
@@ -300,9 +331,11 @@ type_validators.table = function (schema : json_schema_object, data : any, root
 		end
 
 		local p = 0
-		if schema.prefixItems then
+		if schema.prefixItems ~= nil then
 			for i, s in ipairs(schema.prefixItems) do
-				if validate(s, data[i], root) then
+				if data[i] == nil then
+					break
+				elseif validate(s, data[i], root) then
 					p = i
 				else
 					return false
@@ -310,7 +343,7 @@ type_validators.table = function (schema : json_schema_object, data : any, root
 			end
 		end
 
-		if schema.items then
+		if schema.items ~= nil then
 			for i = p+1, #data do
 				if not validate(schema.items, data[i], root) then
 					return false
@@ -318,7 +351,7 @@ type_validators.table = function (schema : json_schema_object, data : any, root
 			end
 		end
 
-		if schema.contains then
+		if schema.contains ~= nil then
 			local found = false
 			for i = 1, #data do
 				if validate(schema.contains, data[i], root) then
@@ -330,39 +363,11 @@ type_validators.table = function (schema : json_schema_object, data : any, root
 				return false
 			end
 		end
-
-		return true
 	end
-	return false
+
+	return true;
 end
 
-type_validators.object = function (schema : schema_t, data : any, root : json_schema_object) : boolean
-	if data is table then
-		for k in pairs(data) do
-			if not k is string then
-				return false
-			end
-		end
-
-		return type_validators.table(schema, data, root)
-	end
-	return false
-end
-
-type_validators.array = function (schema : schema_t, data : any, root : json_schema_object) : boolean
-	if data is table then
-
-		-- just check that there the keys are all numbers
-		for i in pairs(data) do
-			if not i is number then
-				return false
-			end
-		end
-
-		return type_validators.table(schema, data, root)
-	end
-	return false
-end
 
 json_schema_object.validate = validate;
 

util/jsonschema.lua

@@ -1,3 +1,5 @@
+-- This file is generated from teal-src/util/jsonschema.lua
+
 local m_type = math.type or function (n)
 	return n % 1 == 0 and n <= 9007199254740992 and n >= -9007199254740992 and "integer" or "float";
 end;
@@ -10,12 +12,12 @@ local json_type_name = json.json_type_name
 
 local schema_t = {}
 
-local json_schema_object = {xml_t = {}}
+local json_schema_object = { xml_t = {} }
-
-local type_validators = {}
 
 local function simple_validate(schema, data)
-	if schema == "object" and type(data) == "table" then
+	if schema == nil then
+		return true
+	elseif schema == "object" and type(data) == "table" then
 		return type(data) == "table" and (next(data) == nil or type((next(data, nil))) == "string")
 	elseif schema == "array" and type(data) == "table" then
 		return type(data) == "table" and (next(data) == nil or type((next(data, nil))) == "number")
@@ -23,12 +25,82 @@ local function simple_validate(schema, data)
 		return m_type(data) == schema
 	elseif schema == "null" then
 		return data == null
+	elseif type(schema) == "table" then
+		for _, one in ipairs(schema) do
+			if simple_validate(one, data) then
+				return true
+			end
+		end
+		return false
 	else
 		return type(data) == schema
 	end
 end
 
-type_validators.string = function(schema, data)
+local complex_validate
+
+local function validate(schema, data, root)
+	if type(schema) == "boolean" then
+		return schema
+	else
+		return complex_validate(schema, data, root)
+	end
+end
+
+function complex_validate(schema, data, root)
+
+	if root == nil then
+		root = schema
+	end
+
+	if schema["$ref"] and schema["$ref"]:sub(1, 1) == "#" then
+		local referenced = pointer.resolve(root, schema["$ref"]:sub(2))
+		if referenced ~= nil and referenced ~= root and referenced ~= schema then
+			if not validate(referenced, data, root) then
+				return false
+			end
+		end
+	end
+
+	if not simple_validate(schema.type, data) then
+		return false
+	end
+
+	if schema.type == "object" then
+		if type(data) == "table" then
+
+			for k in pairs(data) do
+				if not (type(k) == "string") then
+					return false
+				end
+			end
+		end
+	end
+
+	if schema.type == "array" then
+		if type(data) == "table" then
+
+			for i in pairs(data) do
+				if not (math.type(i) == "integer") then
+					return false
+				end
+			end
+		end
+	end
+
+	if schema["enum"] ~= nil then
+		local match = false
+		for _, v in ipairs(schema["enum"]) do
+			if v == data then
+
+				match = true
+				break
+			end
+		end
+		if not match then
+			return false
+		end
+	end
 
 	if type(data) == "string" then
 		if schema.maxLength and #data > schema.maxLength then
@@ -37,13 +109,10 @@ type_validators.string = function(schema, data)
 		if schema.minLength and #data < schema.minLength then
 			return false
 		end
-		return true
 	end
-	return false
-end
 
-type_validators.number = function(schema, data)
+	if type(data) == "number" then
-	if schema.multipleOf and data % schema.multipleOf ~= 0 then
+		if schema.multipleOf and (data == 0 or data % schema.multipleOf ~= 0) then
 			return false
 		end
 
@@ -62,28 +131,6 @@ type_validators.number = function(schema, data)
 		if schema.exclusiveMinimum and not (data > schema.exclusiveMinimum) then
 			return false
 		end
-
-	return true
-end
-
-type_validators.integer = type_validators.number
-
-local function validate(schema, data, root)
-	if type(schema) == "boolean" then
-		return schema
-	end
-	if type(schema) == "string" then
-		return simple_validate(schema, data)
-	end
-	if type(schema) == "table" then
-		if root == nil then
-			root = schema
-		end
-		if schema["$ref"] and schema["$ref"]:sub(1, 1) == "#" then
-			local referenced = pointer.resolve(root, schema["$ref"]:sub(2))
-			if referenced ~= nil then
-				return validate(referenced, data, root)
-			end
 	end
 
 	if schema.allOf then
@@ -92,7 +139,6 @@ local function validate(schema, data, root)
 				return false
 			end
 		end
-			return true
 	end
 
 	if schema.oneOf then
@@ -102,17 +148,23 @@ local function validate(schema, data, root)
 				valid = valid + 1
 			end
 		end
-			return valid == 1
+		if valid ~= 1 then
+			return false
+		end
 	end
 
 	if schema.anyOf then
+		local match = false
 		for _, sub in ipairs(schema.anyOf) do
 			if validate(sub, data, root) then
-					return true
+				match = true
+				break
 			end
 		end
+		if not match then
 			return false
 		end
+	end
 
 	if schema["not"] then
 		if validate(schema["not"], data, root) then
@@ -120,7 +172,7 @@ local function validate(schema, data, root)
 		end
 	end
 
-		if schema["if"] then
+	if schema["if"] ~= nil then
 		if validate(schema["if"], data, root) then
 			if schema["then"] then
 				return validate(schema["then"], data, root)
@@ -136,30 +188,6 @@ local function validate(schema, data, root)
 		return false
 	end
 
-		if schema["enum"] ~= nil then
-			for _, v in ipairs(schema["enum"]) do
-				if v == data then
-					return true
-				end
-			end
-			return false
-		end
-
-		if schema.type then
-			if not simple_validate(schema.type, data) then
-				return false
-			end
-
-			local validator = type_validators[schema.type]
-			if validator then
-				return validator(schema, data, root)
-			end
-		end
-		return true
-	end
-end
-
-type_validators.table = function(schema, data, root)
 	if type(data) == "table" then
 
 		if schema.maxItems and #data > schema.maxItems then
@@ -178,27 +206,31 @@ type_validators.table = function(schema, data, root)
 			end
 		end
 
+		if schema.propertyNames ~= nil then
+			for k in pairs(data) do
+				if not validate(schema.propertyNames, k, root) then
+					return false
+				end
+			end
+		end
+
 		if schema.properties then
-			local additional = schema.additionalProperties or true
+			for k, sub in pairs(schema.properties) do
+				if data[k] ~= nil and not validate(sub, data[k], root) then
+					return false
+				end
+			end
+		end
+
+		if schema.additionalProperties ~= nil then
 			for k, v in pairs(data) do
-				if schema.propertyNames and not validate(schema.propertyNames, k, root) then
+				if schema.properties == nil or schema.properties[k] == nil then
-					return false
-				end
-				local s = schema.properties[k] or additional
-				if not validate(s, v, root) then
-					return false
-				end
-			end
-		elseif schema.additionalProperties then
-			for k, v in pairs(data) do
-				if schema.propertyNames and not validate(schema.propertyNames, k, root) then
-					return false
-				end
 					if not validate(schema.additionalProperties, v, root) then
 						return false
 					end
 				end
 			end
+		end
 
 		if schema.uniqueItems then
 
@@ -212,9 +244,11 @@ type_validators.table = function(schema, data, root)
 		end
 
 		local p = 0
-		if schema.prefixItems then
+		if schema.prefixItems ~= nil then
 			for i, s in ipairs(schema.prefixItems) do
-				if validate(s, data[i], root) then
+				if data[i] == nil then
+					break
+				elseif validate(s, data[i], root) then
 					p = i
 				else
 					return false
@@ -222,7 +256,7 @@ type_validators.table = function(schema, data, root)
 			end
 		end
 
-		if schema.items then
+		if schema.items ~= nil then
 			for i = p + 1, #data do
 				if not validate(schema.items, data[i], root) then
 					return false
@@ -230,7 +264,7 @@ type_validators.table = function(schema, data, root)
 			end
 		end
 
-		if schema.contains then
+		if schema.contains ~= nil then
 			local found = false
 			for i = 1, #data do
 				if validate(schema.contains, data[i], root) then
@@ -242,37 +276,9 @@ type_validators.table = function(schema, data, root)
 				return false
 			end
 		end
+	end
 
 	return true
-	end
-	return false
-end
-
-type_validators.object = function(schema, data, root)
-	if type(data) == "table" then
-		for k in pairs(data) do
-			if not (type(k) == "string") then
-				return false
-			end
-		end
-
-		return type_validators.table(schema, data, root)
-	end
-	return false
-end
-
-type_validators.array = function(schema, data, root)
-	if type(data) == "table" then
-
-		for i in pairs(data) do
-			if not (type(i) == "number") then
-				return false
-			end
-		end
-
-		return type_validators.table(schema, data, root)
-	end
-	return false
 end
 
 json_schema_object.validate = validate;