Merge branch 'json-schema-fso' into HEAD

doc/manual/package.nix

@@ -34,6 +34,7 @@ mkMesonDerivation (finalAttrs: {
       (fileset.unions [
         ../../.version
         # For example JSON
+        ../../src/libutil-tests/data/memory-source-accessor
         ../../src/libutil-tests/data/hash
         ../../src/libstore-tests/data/content-address
         # Too many different types of files to filter for now

doc/manual/source/SUMMARY.md.in

@@ -117,6 +117,7 @@
 - [Architecture and Design](architecture/architecture.md)
 - [Formats and Protocols](protocols/index.md)
   - [JSON Formats](protocols/json/index.md)
+    - [File System Object](protocols/json/file-system-object.md)
     - [Hash](protocols/json/hash.md)
     - [Pseudo Content Address](protocols/json/content-address.md)
     - [Store Object Info](protocols/json/store-object-info.md)

doc/manual/source/protocols/json/file-system-object.md

@@ -0,0 +1,21 @@
+{{#include file-system-object-v1-fixed.md}}
+
+## Examples
+
+### Simple
+
+```json
+{{#include schema/file-system-object-v1/simple.json}}
+```
+
+### Complex
+
+```json
+{{#include schema/file-system-object-v1/complex.json}}
+```
+
+<!--
+## Raw Schema
+
+[JSON Schema for File System Object v1](schema/file-system-object-v1.json)
+-->

doc/manual/source/protocols/json/fixup-json-schema-generated-doc.sed

@@ -11,6 +11,7 @@ s/\\`/`/g
 #
 # As we have more such relative links, more replacements of this nature
 # should appear below.
+s^#/\$defs/\(regular\|symlink\|directory\)^In this schema^g
 s^\(./hash-v1.yaml\)\?#/$defs/algorithm^[JSON format for `Hash`](./hash.html#algorithm)^g
 s^\(./hash-v1.yaml\)^[JSON format for `Hash`](./hash.html)^g
 s^\(./content-address-v1.yaml\)\?#/$defs/method^[JSON format for `ContentAddress`](./content-address.html#method)^g

doc/manual/source/protocols/json/meson.build

@@ -9,6 +9,7 @@ json_schema_for_humans = find_program('generate-schema-doc', required : false)
 json_schema_config = files('json-schema-for-humans-config.yaml')
 
 schemas = [
+  'file-system-object-v1',
   'hash-v1',
   'content-address-v1',
   'derivation-v3',

doc/manual/source/protocols/json/schema/file-system-object-v1

@@ -0,0 +1 @@
+../../../../../../src/libutil-tests/data/memory-source-accessor

doc/manual/source/protocols/json/schema/file-system-object-v1.yaml

@@ -0,0 +1,65 @@
+"$schema": http://json-schema.org/draft-04/schema#
+"$id": https://nix.dev/manual/nix/latest/protocols/json/schema/file-system-object-v1.json
+title: File System Object
+description: |
+  This schema describes the JSON representation of Nix's [File System Object](@docroot@/store/file-system-object.md).
+
+  The schema is recursive because file system objects contain other file system objects.
+type: object
+required: ["type"]
+properties:
+  type:
+    type: string
+    enum: ["regular", "symlink", "directory"]
+
+# Enforce conditional structure based on `type`
+anyOf:
+  - $ref: "#/$defs/regular"
+    required: ["type", "contents"]
+
+  - $ref: "#/$defs/symlink"
+    required: ["type", "target"]
+
+  - $ref: "#/$defs/directory"
+    required: ["type", "contents"]
+
+"$defs":
+  regular:
+    title: Regular File
+    required: ["contents"]
+    properties:
+      type:
+        const: "regular"
+      contents:
+        type: string
+        description: Base64-encoded file contents
+      executable:
+        type: boolean
+        description: Whether the file is executable.
+        default: false
+    additionalProperties: false
+
+  symlink:
+    title: Symbolic Link
+    required: ["target"]
+    properties:
+      type:
+        const: "symlink"
+      target:
+        type: string
+        description: Target path of the symlink.
+    additionalProperties: false
+
+  directory:
+    title: Directory
+    required: ["contents"]
+    properties:
+      type:
+        const: "directory"
+      contents:
+        type: object
+        description: |
+          Map of names to nested file system objects (for type=directory)
+        additionalProperties:
+          $ref: "#"
+    additionalProperties: false

src/json-schema-checks/file-system-object

@@ -0,0 +1 @@
+../../src/libutil-tests/data/memory-source-accessor

src/json-schema-checks/meson.build

@@ -20,6 +20,14 @@ schema_dir = meson.current_source_dir() / 'schema'
 
 # Get all example files
 schemas = [
+  {
+    'stem' : 'file-system-object',
+    'schema' : schema_dir / 'file-system-object-v1.yaml',
+    'files' : [
+      'simple.json',
+      'complex.json',
+    ],
+  },
   {
     'stem' : 'hash',
     'schema' : schema_dir / 'hash-v1.yaml',

src/json-schema-checks/package.nix

@@ -20,6 +20,7 @@ mkMesonDerivation (finalAttrs: {
   fileset = lib.fileset.unions [
     ../../.version
     ../../doc/manual/source/protocols/json/schema
+    ../../src/libutil-tests/data/memory-source-accessor
     ../../src/libutil-tests/data/hash
     ../../src/libstore-tests/data/content-address
     ../../src/libstore-tests/data/derivation

src/libutil-tests/data/memory-source-accessor/complex.json

@@ -0,0 +1,24 @@
+{
+  "contents": {
+    "bar": {
+      "contents": {
+        "baz": {
+          "contents": "Z29vZCBkYXksCg==",
+          "executable": true,
+          "type": "regular"
+        },
+        "quux": {
+          "target": "/over/there",
+          "type": "symlink"
+        }
+      },
+      "type": "directory"
+    },
+    "foo": {
+      "contents": "aGVsbG8K",
+      "executable": false,
+      "type": "regular"
+    }
+  },
+  "type": "directory"
+}

src/libutil-tests/data/memory-source-accessor/simple.json

@@ -0,0 +1,5 @@
+{
+  "contents": "YXNkZg==",
+  "executable": false,
+  "type": "regular"
+}

src/libutil-tests/git.cc

@@ -224,42 +224,15 @@ TEST_F(GitTest, tree_sha256_write)
     });
 }
 
+namespace memory_source_accessor {
+
+extern ref<MemorySourceAccessor> exampleComplex();
+
+}
+
 TEST_F(GitTest, both_roundrip)
 {
-    using File = MemorySourceAccessor::File;
+    auto files = memory_source_accessor::exampleComplex();
-
-    auto files = make_ref<MemorySourceAccessor>();
-    files->root = File::Directory{
-        .contents{
-            {
-                "foo",
-                File::Regular{
-                    .contents = "hello\n\0\n\tworld!",
-                },
-            },
-            {
-                "bar",
-                File::Directory{
-                    .contents =
-                        {
-                            {
-                                "baz",
-                                File::Regular{
-                                    .executable = true,
-                                    .contents = "good day,\n\0\n\tworld!",
-                                },
-                            },
-                            {
-                                "quux",
-                                File::Symlink{
-                                    .target = "/over/there",
-                                },
-                            },
-                        },
-                },
-            },
-        },
-    };
 
     for (const auto hashAlgo : {HashAlgorithm::SHA1, HashAlgorithm::SHA256}) {
         std::map<Hash, std::string> cas;

src/libutil-tests/memory-source-accessor.cc

@@ -0,0 +1,116 @@
+#include <string_view>
+
+#include "nix/util/bytes.hh"
+#include "nix/util/memory-source-accessor.hh"
+#include "nix/util/tests/json-characterization.hh"
+
+namespace nix {
+
+namespace memory_source_accessor {
+
+using File = MemorySourceAccessor::File;
+
+ref<MemorySourceAccessor> exampleSimple()
+{
+    auto sc = make_ref<MemorySourceAccessor>();
+    sc->root = File{File::Regular{
+        .executable = false,
+        .contents = to_owned(as_bytes("asdf")),
+    }};
+    return sc;
+}
+
+ref<MemorySourceAccessor> exampleComplex()
+{
+    auto files = make_ref<MemorySourceAccessor>();
+    files->root = File::Directory{
+        .contents{
+            {
+                "foo",
+                File::Regular{
+                    .contents = to_owned(as_bytes("hello\n\0\n\tworld!")),
+                },
+            },
+            {
+                "bar",
+                File::Directory{
+                    .contents =
+                        {
+                            {
+                                "baz",
+                                File::Regular{
+                                    .executable = true,
+                                    .contents = to_owned(as_bytes("good day,\n\0\n\tworld!")),
+                                },
+                            },
+                            {
+                                "quux",
+                                File::Symlink{
+                                    .target = "/over/there",
+                                },
+                            },
+                        },
+                },
+            },
+        },
+    };
+    return files;
+}
+
+} // namespace memory_source_accessor
+
+/* ----------------------------------------------------------------------------
+ * JSON
+ * --------------------------------------------------------------------------*/
+
+class MemorySourceAccessorTest : public virtual CharacterizationTest
+{
+    std::filesystem::path unitTestData = getUnitTestData() / "memory-source-accessor";
+
+public:
+
+    std::filesystem::path goldenMaster(std::string_view testStem) const override
+    {
+        return unitTestData / testStem;
+    }
+};
+
+using nlohmann::json;
+
+struct MemorySourceAccessorJsonTest : MemorySourceAccessorTest,
+                                      JsonCharacterizationTest<MemorySourceAccessor>,
+                                      ::testing::WithParamInterface<std::pair<std::string_view, MemorySourceAccessor>>
+{};
+
+TEST_P(MemorySourceAccessorJsonTest, from_json)
+{
+    auto & [name, expected] = GetParam();
+    /* Cannot use `readJsonTest` because need to compare `root` field of
+       the source accessors for equality. */
+    readTest(Path{name} + ".json", [&](const auto & encodedRaw) {
+        auto encoded = json::parse(encodedRaw);
+        auto decoded = static_cast<MemorySourceAccessor>(encoded);
+        ASSERT_EQ(decoded.root, expected.root);
+    });
+}
+
+TEST_P(MemorySourceAccessorJsonTest, to_json)
+{
+    auto & [name, value] = GetParam();
+    writeJsonTest(name, value);
+}
+
+INSTANTIATE_TEST_SUITE_P(
+    MemorySourceAccessorJSON,
+    MemorySourceAccessorJsonTest,
+    ::testing::Values(
+        std::pair{
+            "simple",
+            *memory_source_accessor::exampleSimple(),
+        },
+        std::pair{
+            "complex",
+            *memory_source_accessor::exampleComplex(),
+        }));
+
+} // namespace nix

src/libutil-tests/meson.build

@@ -63,6 +63,7 @@ sources = files(
   'json-utils.cc',
   'logging.cc',
   'lru-cache.cc',
+  'memory-source-accessor.cc',
   'monitorfdhup.cc',
   'nix_api_util.cc',
   'nix_api_util_internal.cc',

src/libutil/include/nix/util/bytes.hh

@@ -0,0 +1,41 @@
+#pragma once
+///@file
+
+#include <string_view>
+#include <span>
+#include <vector>
+
+namespace nix {
+
+static inline std::span<const std::byte> as_bytes(std::string_view sv) noexcept
+{
+    return std::span<const std::byte>{
+        reinterpret_cast<const std::byte *>(sv.data()),
+        sv.size(),
+    };
+}
+
+static inline std::vector<std::byte> to_owned(std::span<const std::byte> bytes)
+{
+    return std::vector<std::byte>{
+        bytes.begin(),
+        bytes.end(),
+    };
+}
+
+/**
+ * @note this should be avoided, as arbitrary binary data in strings
+ * views, while allowed, is not really proper. Generally this should
+ * only be used as a stop-gap with other definitions that themselves
+ * should be converted to accept `std::span<const std::byte>` or
+ * similar, directly.
+ */
+static inline std::string_view to_str(std::span<const std::byte> sp)
+{
+    return std::string_view{
+        reinterpret_cast<const char *>(sp.data()),
+        sp.size(),
+    };
+}
+
+} // namespace nix

src/libutil/include/nix/util/memory-source-accessor.hh

@@ -4,6 +4,7 @@
 #include "nix/util/source-path.hh"
 #include "nix/util/fs-sink.hh"
 #include "nix/util/variant-wrapper.hh"
+#include "nix/util/json-impls.hh"
 
 namespace nix {
 
@@ -25,7 +26,7 @@ struct MemorySourceAccessor : virtual SourceAccessor
         struct Regular
         {
             bool executable = false;
-            std::string contents;
+            std::vector<std::byte> contents;
 
             bool operator==(const Regular &) const = default;
             auto operator<=>(const Regular &) const = default;
@@ -86,7 +87,13 @@ struct MemorySourceAccessor : virtual SourceAccessor
      */
     File * open(const CanonPath & path, std::optional<File> create);
 
-    SourcePath addFile(CanonPath path, std::string && contents);
+    SourcePath addFile(CanonPath path, std::vector<std::byte> && contents);
+
+    /**
+     * Small wrapper of the other `addFile`, purely for convenience when
+     * the file in question to be added is a string.
+     */
+    SourcePath addFile(CanonPath path, std::string_view contents);
 };
 
 inline bool MemorySourceAccessor::File::Directory::operator==(
@@ -121,4 +128,30 @@ struct MemorySink : FileSystemObjectSink
     void createSymlink(const CanonPath & path, const std::string & target) override;
 };
 
+template<>
+struct json_avoids_null<MemorySourceAccessor::File::Regular> : std::true_type
+{};
+
+template<>
+struct json_avoids_null<MemorySourceAccessor::File::Directory> : std::true_type
+{};
+
+template<>
+struct json_avoids_null<MemorySourceAccessor::File::Symlink> : std::true_type
+{};
+
+template<>
+struct json_avoids_null<MemorySourceAccessor::File> : std::true_type
+{};
+
+template<>
+struct json_avoids_null<MemorySourceAccessor> : std::true_type
+{};
+
 } // namespace nix
+
+JSON_IMPL(MemorySourceAccessor::File::Regular)
+JSON_IMPL(MemorySourceAccessor::File::Directory)
+JSON_IMPL(MemorySourceAccessor::File::Symlink)
+JSON_IMPL(MemorySourceAccessor::File)
+JSON_IMPL(MemorySourceAccessor)

src/libutil/include/nix/util/meson.build

@@ -12,6 +12,7 @@ headers = files(
   'array-from-string-literal.hh',
   'base-n.hh',
   'base-nix-32.hh',
+  'bytes.hh',
   'callback.hh',
   'canon-path.hh',
   'checked-arithmetic.hh',

src/libutil/memory-source-accessor.cc

@@ -1,4 +1,7 @@
 #include "nix/util/memory-source-accessor.hh"
+#include "nix/util/base-n.hh"
+#include "nix/util/bytes.hh"
+#include "nix/util/json-utils.hh"
 
 namespace nix {
 
@@ -58,7 +61,7 @@ std::string MemorySourceAccessor::readFile(const CanonPath & path)
     if (!f)
         throw Error("file '%s' does not exist", path);
     if (auto * r = std::get_if<File::Regular>(&f->raw))
-        return r->contents;
+        return std::string{to_str(r->contents)};
     else
         throw Error("file '%s' is not a regular file", path);
 }
@@ -125,7 +128,7 @@ std::string MemorySourceAccessor::readLink(const CanonPath & path)
         throw Error("file '%s' is not a symbolic link", path);
 }
 
-SourcePath MemorySourceAccessor::addFile(CanonPath path, std::string && contents)
+SourcePath MemorySourceAccessor::addFile(CanonPath path, std::vector<std::byte> && contents)
 {
     // Create root directory automatically if necessary as a convenience.
     if (!root && !path.isRoot())
@@ -142,6 +145,11 @@ SourcePath MemorySourceAccessor::addFile(CanonPath path, std::string && contents
     return SourcePath{ref(shared_from_this()), path};
 }
 
+SourcePath MemorySourceAccessor::addFile(CanonPath path, std::string_view contents)
+{
+    return addFile(path, to_owned(as_bytes(contents)));
+}
+
 using File = MemorySourceAccessor::File;
 
 void MemorySink::createDirectory(const CanonPath & path)
@@ -190,9 +198,10 @@ void CreateMemoryRegularFile::preallocateContents(uint64_t len)
     regularFile.contents.reserve(len);
 }
 
-void CreateMemoryRegularFile::operator()(std::string_view data)
+void CreateMemoryRegularFile::operator()(std::string_view data_)
 {
-    regularFile.contents += data;
+    auto data = as_bytes(data_);
+    regularFile.contents.insert(regularFile.contents.end(), data.begin(), data.end());
 }
 
 void MemorySink::createSymlink(const CanonPath & path, const std::string & target)
@@ -222,3 +231,106 @@ ref<SourceAccessor> makeEmptySourceAccessor()
 }
 
 } // namespace nix
+
+namespace nlohmann {
+
+using namespace nix;
+
+MemorySourceAccessor::File::Regular adl_serializer<MemorySourceAccessor::File::Regular>::from_json(const json & json)
+{
+    auto & obj = getObject(json);
+    return MemorySourceAccessor::File::Regular{
+        .executable = getBoolean(valueAt(obj, "executable")),
+        .contents = to_owned(as_bytes(base64::decode(getString(valueAt(obj, "contents"))))),
+    };
+}
+
+void adl_serializer<MemorySourceAccessor::File::Regular>::to_json(
+    json & json, const MemorySourceAccessor::File::Regular & val)
+{
+    json = {
+        {"executable", val.executable},
+        {"contents", base64::encode(val.contents)},
+    };
+}
+
+MemorySourceAccessor::File::Directory
+adl_serializer<MemorySourceAccessor::File::Directory>::from_json(const json & json)
+{
+    auto & obj = getObject(json);
+    return MemorySourceAccessor::File::Directory{
+        .contents = valueAt(obj, "contents"),
+    };
+}
+
+void adl_serializer<MemorySourceAccessor::File::Directory>::to_json(
+    json & json, const MemorySourceAccessor::File::Directory & val)
+{
+    json = {
+        {"contents", val.contents},
+    };
+}
+
+MemorySourceAccessor::File::Symlink adl_serializer<MemorySourceAccessor::File::Symlink>::from_json(const json & json)
+{
+    auto & obj = getObject(json);
+    return MemorySourceAccessor::File::Symlink{
+        .target = getString(valueAt(obj, "target")),
+    };
+}
+
+void adl_serializer<MemorySourceAccessor::File::Symlink>::to_json(
+    json & json, const MemorySourceAccessor::File::Symlink & val)
+{
+    json = {
+        {"target", val.target},
+    };
+}
+
+MemorySourceAccessor::File adl_serializer<MemorySourceAccessor::File>::from_json(const json & json)
+{
+    auto & obj = getObject(json);
+    auto type = getString(valueAt(obj, "type"));
+    if (type == "regular")
+        return static_cast<MemorySourceAccessor::File::Regular>(json);
+    if (type == "directory")
+        return static_cast<MemorySourceAccessor::File::Directory>(json);
+    if (type == "symlink")
+        return static_cast<MemorySourceAccessor::File::Symlink>(json);
+    else
+        throw Error("unknown type of file '%s'", type);
+}
+
+void adl_serializer<MemorySourceAccessor::File>::to_json(json & json, const MemorySourceAccessor::File & val)
+{
+    std::visit(
+        overloaded{
+            [&](const MemorySourceAccessor::File::Regular & r) {
+                json = r;
+                json["type"] = "regular";
+            },
+            [&](const MemorySourceAccessor::File::Directory & d) {
+                json = d;
+                json["type"] = "directory";
+            },
+            [&](const MemorySourceAccessor::File::Symlink & s) {
+                json = s;
+                json["type"] = "symlink";
+            },
+        },
+        val.raw);
+}
+
+MemorySourceAccessor adl_serializer<MemorySourceAccessor>::from_json(const json & json)
+{
+    MemorySourceAccessor res;
+    res.root = json;
+    return res;
+}
+
+void adl_serializer<MemorySourceAccessor>::to_json(json & json, const MemorySourceAccessor & val)
+{
+    json = val.root;
+}
+
+} // namespace nlohmann