Expose JSON BinPack's canonicalizer as a canonicalize command

Signed-off-by: Juan Cruz Viotti <[email protected]>

CMakeLists.txt

@@ -11,6 +11,7 @@ option(JSONSCHEMA_CONTINUOUS "Perform a continuous JSON Schema CLI release" ON)
 
 find_package(JSONToolkit REQUIRED)
 find_package(AlterSchema REQUIRED)
+find_package(JSONBinPack REQUIRED)
 find_package(Hydra REQUIRED)
 add_subdirectory(src)
 

DEPENDENCIES

@@ -3,3 +3,4 @@ noa https://github.com/sourcemeta/noa 7e26abce7a4e31e86a16ef2851702a56773ca527
 jsontoolkit https://github.com/sourcemeta/jsontoolkit 3ef19daf7ca042544239111c701a51232f3f5576
 hydra https://github.com/sourcemeta/hydra 3c53d3fdef79e9ba603d48470a508cc45472a0dc
 alterschema https://github.com/sourcemeta/alterschema 744cf03a950b681a61f1f4cf6a7bb55bc52836c9
+jsonbinpack https://github.com/sourcemeta/jsonbinpack 43d53dd32c432333deb1aea147095ed8707b5f11

README.markdown

@@ -65,6 +65,7 @@ documentation:
 - [`jsonschema frame`](./docs/frame.markdown) (for debugging references)
 - [`jsonschema compile`](./docs/compile.markdown) (for internal debugging)
 - [`jsonschema identify`](./docs/identify.markdown)
+- [`jsonschema canonicalize`](./docs/canonicalize.markdown) (for static analysis)
 
 Installation
 ------------

cmake/FindJSONBinPack.cmake

@@ -0,0 +1,7 @@
+if(NOT JSONBinPack_FOUND)
+  set(JSONBINPACK_INSTALL OFF CACHE BOOL "disable installation")
+  set(JSONBINPACK_CLI OFF CACHE BOOL "disable the JSON BinPack CLI module")
+  set(JSONBINPACK_RUNTIME OFF CACHE BOOL "disable the JSON BinPack runtime module")
+  add_subdirectory("${PROJECT_SOURCE_DIR}/vendor/jsonbinpack")
+  set(JSONBinPack_FOUND ON)
+endif()

docs/canonicalize.markdown

@@ -0,0 +1,83 @@
+Canonicalize
+============
+
+```sh
+jsonschema canonicalize <schema.json>
+```
+
+JSON Schema is an extremely expressive schema language. As such, schema authors
+can express the same constraints in a variety of ways, making the process of
+statically analyzing schemas complex. This command attempts to tackle the
+problem by transforming a given JSON Schema into a simpler (but more verbose)
+normalized form referred to as _canonical_.
+
+> Refer to [Juan Cruz Viotti's dissertation on JSON
+> BinPack's](https://www.jviotti.com/dissertation.pdf) for how JSON Schema
+> canonicalization was originally defined.
+
+Examples
+--------
+
+For example, consider the following simple schema:
+
+```json
+{
+  "$schema": "https://json-schema.org/draft/2020-12/schema",
+  "properties": {
+    "foo": { "type": "string" }
+  }
+}
+```
+
+The canonicalization process will result in something like this:
+
+```
+{
+  "$schema": "https://json-schema.org/draft/2020-12/schema",
+  "anyOf": [
+    {
+      "enum": [
+        null
+      ]
+    },
+    {
+      "enum": [
+        false,
+        true
+      ]
+    },
+    {
+      "type": "object",
+      "minProperties": 0,
+      "properties": {
+        "foo": {
+          "type": "string",
+          "minLength": 0
+        }
+      }
+    },
+    {
+      "type": "array",
+      "minItems": 0
+    },
+    {
+      "type": "string",
+      "minLength": 0
+    },
+    {
+      "type": "number",
+      "multipleOf": 1
+    },
+    {
+      "type": "integer",
+      "multipleOf": 1
+    }
+  ]
+}
+```
+
+### Canonicalize a JSON Schema
+
+```sh
+jsonschema canonicalize path/to/my/schema.json
+```

src/CMakeLists.txt

@@ -9,7 +9,8 @@ add_executable(jsonschema_cli
   command_lint.cc
   command_metaschema.cc
   command_validate.cc
-  command_identify.cc)
+  command_identify.cc
+  command_canonicalize.cc)
 
 noa_add_default_options(PRIVATE jsonschema_cli)
 set_target_properties(jsonschema_cli PROPERTIES OUTPUT_NAME jsonschema)
@@ -20,6 +21,7 @@ target_link_libraries(jsonschema_cli PRIVATE sourcemeta::jsontoolkit::jsonschema
 target_link_libraries(jsonschema_cli PRIVATE sourcemeta::alterschema::engine)
 target_link_libraries(jsonschema_cli PRIVATE sourcemeta::alterschema::linter)
 target_link_libraries(jsonschema_cli PRIVATE sourcemeta::hydra::httpclient)
+target_link_libraries(jsonschema_cli PRIVATE sourcemeta::jsonbinpack::compiler)
 
 configure_file(configure.h.in configure.h @ONLY)
 target_include_directories(jsonschema_cli PRIVATE "${CMAKE_CURRENT_BINARY_DIR}")

src/command.h

@@ -14,6 +14,7 @@ auto lint(const std::span<const std::string> &arguments) -> int;
 auto validate(const std::span<const std::string> &arguments) -> int;
 auto metaschema(const std::span<const std::string> &arguments) -> int;
 auto identify(const std::span<const std::string> &arguments) -> int;
+auto canonicalize(const std::span<const std::string> &arguments) -> int;
 } // namespace sourcemeta::jsonschema::cli
 
 #endif

src/command_canonicalize.cc

@@ -0,0 +1,30 @@
+#include <sourcemeta/jsonbinpack/compiler.h>
+#include <sourcemeta/jsontoolkit/json.h>
+#include <sourcemeta/jsontoolkit/jsonschema.h>
+
+#include <cstdlib>  // EXIT_SUCCESS
+#include <iostream> // std::cout, std::endl
+
+#include "command.h"
+#include "utils.h"
+
+auto sourcemeta::jsonschema::cli::canonicalize(
+    const std::span<const std::string> &arguments) -> int {
+  const auto options{parse_options(arguments, {})};
+
+  if (options.at("").size() < 1) {
+    std::cerr
+        << "error: This command expects a path to a schema. For example:\n\n"
+        << "  jsonschema canonicalize path/to/schema.json\n";
+    return EXIT_FAILURE;
+  }
+
+  auto schema{sourcemeta::jsontoolkit::from_file(options.at("").front())};
+  sourcemeta::jsonbinpack::canonicalize(
+      schema, sourcemeta::jsontoolkit::default_schema_walker,
+      resolver(options, options.contains("h") || options.contains("http")));
+  sourcemeta::jsontoolkit::prettify(
+      schema, std::cout, sourcemeta::jsontoolkit::schema_format_compare);
+  std::cout << std::endl;
+  return EXIT_SUCCESS;
+}

src/main.cc

@@ -69,6 +69,11 @@ Global Options:
        Print the URI of the given schema to standard output, optionally
        relative to a given base URI.
 
+   canonicalize <schema.json>
+
+       Pre-process a JSON Schema into JSON BinPack's canonical form
+       for static analysis.
+
 For more documentation, visit https://github.com/sourcemeta/jsonschema
 )EOF"};
 
@@ -92,6 +97,8 @@ auto jsonschema_main(const std::string &program, const std::string &command,
     return sourcemeta::jsonschema::cli::test(arguments);
   } else if (command == "identify") {
     return sourcemeta::jsonschema::cli::identify(arguments);
+  } else if (command == "canonicalize") {
+    return sourcemeta::jsonschema::cli::canonicalize(arguments);
   } else {
     std::cout << "JSON Schema CLI - v"
               << sourcemeta::jsonschema::cli::PROJECT_VERSION << "\n";

test/CMakeLists.txt

@@ -166,6 +166,12 @@ add_jsonschema_test_unix(lint/pass_lint_fix)
 add_jsonschema_test_unix(lint/pass_lint_no_fix)
 add_jsonschema_test_unix(lint/fail_lint)
 
+# Canonicalize
+add_jsonschema_test_unix(canonicalize/pass_1)
+add_jsonschema_test_unix(canonicalize/fail_no_schema)
+add_jsonschema_test_unix(canonicalize/fail_schema_invalid_json)
+add_jsonschema_test_unix(canonicalize/fail_unknown_metaschema)
+
 # CI specific tests
 add_jsonschema_test_unix_ci(pass_bundle_http)
 add_jsonschema_test_unix_ci(fail_bundle_http_non_200)

test/canonicalize/fail_no_schema.sh

@@ -0,0 +1,19 @@
+#!/bin/sh
+
+set -o errexit
+set -o nounset
+
+TMP="$(mktemp -d)"
+clean() { rm -rf "$TMP"; }
+trap clean EXIT
+
+"$1" canonicalize 2>"$TMP/stderr.txt" && CODE="$?" || CODE="$?"
+test "$CODE" = "1" || exit 1
+
+cat << 'EOF' > "$TMP/expected.txt"
+error: This command expects a path to a schema. For example:
+
+  jsonschema canonicalize path/to/schema.json
+EOF
+
+diff "$TMP/stderr.txt" "$TMP/expected.txt"

test/canonicalize/fail_schema_invalid_json.sh

@@ -0,0 +1,25 @@
+#!/bin/sh
+
+set -o errexit
+set -o nounset
+
+TMP="$(mktemp -d)"
+clean() { rm -rf "$TMP"; }
+trap clean EXIT
+
+cat << 'EOF' > "$TMP/schema.json"
+{
+  "type" string
+}
+EOF
+
+"$1" canonicalize "$TMP/schema.json" 2>"$TMP/stderr.txt" \
+  && CODE="$?" || CODE="$?"
+test "$CODE" = "1" || exit 1
+
+cat << EOF > "$TMP/expected.txt"
+error: Failed to parse the JSON document at line 2 and column 10
+  $(realpath "$TMP")/schema.json
+EOF
+
+diff "$TMP/stderr.txt" "$TMP/expected.txt"

test/canonicalize/fail_unknown_metaschema.sh

@@ -0,0 +1,26 @@
+#!/bin/sh
+
+set -o errexit
+set -o nounset
+
+TMP="$(mktemp -d)"
+clean() { rm -rf "$TMP"; }
+trap clean EXIT
+
+cat << 'EOF' > "$TMP/schema.json"
+{
+  "$schema": "https://example.com/unknown",
+  "$id": "https://example.com",
+  "$ref": "nested"
+}
+EOF
+
+"$1" canonicalize "$TMP/schema.json" 2>"$TMP/stderr.txt" && CODE="$?" || CODE="$?"
+test "$CODE" = "1" || exit 1
+
+cat << EOF > "$TMP/expected.txt"
+error: Could not resolve the requested schema
+  at https://example.com/unknown
+EOF
+
+diff "$TMP/stderr.txt" "$TMP/expected.txt"

test/canonicalize/pass_1.sh

@@ -0,0 +1,68 @@
+#!/bin/sh
+
+set -o errexit
+set -o nounset
+
+TMP="$(mktemp -d)"
+clean() { rm -rf "$TMP"; }
+trap clean EXIT
+
+cat << 'EOF' > "$TMP/schema.json"
+{
+  "$schema": "https://json-schema.org/draft/2020-12/schema",
+  "properties": {
+    "foo": { "type": "string" }
+  }
+}
+EOF
+
+"$1" canonicalize "$TMP/schema.json" > "$TMP/result.json"
+
+cat "$TMP/result.json"
+
+cat << 'EOF' > "$TMP/expected.json"
+{
+  "$schema": "https://json-schema.org/draft/2020-12/schema",
+  "anyOf": [
+    {
+      "enum": [
+        null
+      ]
+    },
+    {
+      "enum": [
+        false,
+        true
+      ]
+    },
+    {
+      "type": "object",
+      "minProperties": 0,
+      "properties": {
+        "foo": {
+          "type": "string",
+          "minLength": 0
+        }
+      }
+    },
+    {
+      "type": "array",
+      "minItems": 0
+    },
+    {
+      "type": "string",
+      "minLength": 0
+    },
+    {
+      "type": "number",
+      "multipleOf": 1
+    },
+    {
+      "type": "integer",
+      "multipleOf": 1
+    }
+  ]
+}
+EOF
+
+diff "$TMP/result.json" "$TMP/expected.json"