Add Git object hashing to the store layer

Part of RFC 133

Extracted from our old IPFS branches.

Co-Authored-By: Matthew Bauer <mjbauer95@gmail.com>
Co-Authored-By: Carlo Nucera <carlo.nucera@protonmail.com>
Co-authored-by: Robert Hensing <roberth@users.noreply.github.com>
Co-authored-by: Florian Klink <flokli@flokli.de>

src/libutil/file-content-address.cc

@@ -1,16 +1,53 @@
 #include "file-content-address.hh"
 #include "archive.hh"
+#include "git.hh"
 
 namespace nix {
 
-FileIngestionMethod parseFileIngestionMethod(std::string_view input)
+static std::optional<FileSerialisationMethod> parseFileSerialisationMethodOpt(std::string_view input)
 {
     if (input == "flat") {
-        return FileIngestionMethod::Flat;
+        return FileSerialisationMethod::Flat;
     } else if (input == "nar") {
-        return FileIngestionMethod::Recursive;
+        return FileSerialisationMethod::Recursive;
     } else {
-        throw UsageError("Unknown file ingestion method '%s', expect `flat` or `nar`");
+        return std::nullopt;
+    }
+}
+
+FileSerialisationMethod parseFileSerialisationMethod(std::string_view input)
+{
+    auto ret = parseFileSerialisationMethodOpt(input);
+    if (ret)
+        return *ret;
+    else
+        throw UsageError("Unknown file serialiation method '%s', expect `flat` or `nar`");
+}
+
+
+FileIngestionMethod parseFileIngestionMethod(std::string_view input)
+{
+    if (input == "git") {
+        return FileIngestionMethod::Git;
+    } else {
+        auto ret = parseFileSerialisationMethodOpt(input);
+        if (ret)
+            return static_cast<FileIngestionMethod>(*ret);
+        else
+            throw UsageError("Unknown file ingestion method '%s', expect `flat`, `nar`, or `git`");
+    }
+}
+
+
+std::string_view renderFileSerialisationMethod(FileSerialisationMethod method)
+{
+    switch (method) {
+    case FileSerialisationMethod::Flat:
+        return "flat";
+    case FileSerialisationMethod::Recursive:
+        return "nar";
+    default:
+        assert(false);
     }
 }
 
@@ -19,9 +56,11 @@ std::string_view renderFileIngestionMethod(FileIngestionMethod method)
 {
     switch (method) {
     case FileIngestionMethod::Flat:
-        return "flat";
     case FileIngestionMethod::Recursive:
-        return "nar";
+        return renderFileSerialisationMethod(
+            static_cast<FileSerialisationMethod>(method));
+    case FileIngestionMethod::Git:
+        return "git";
     default:
         abort();
     }
@@ -31,14 +70,14 @@ std::string_view renderFileIngestionMethod(FileIngestionMethod method)
 void dumpPath(
     SourceAccessor & accessor, const CanonPath & path,
     Sink & sink,
-    FileIngestionMethod method,
+    FileSerialisationMethod method,
     PathFilter & filter)
 {
     switch (method) {
-    case FileIngestionMethod::Flat:
+    case FileSerialisationMethod::Flat:
         accessor.readFile(path, sink);
         break;
-    case FileIngestionMethod::Recursive:
+    case FileSerialisationMethod::Recursive:
         accessor.dumpPath(path, sink, filter);
         break;
     }
@@ -48,13 +87,13 @@ void dumpPath(
 void restorePath(
     const Path & path,
     Source & source,
-    FileIngestionMethod method)
+    FileSerialisationMethod method)
 {
     switch (method) {
-    case FileIngestionMethod::Flat:
+    case FileSerialisationMethod::Flat:
         writeFile(path, source);
         break;
-    case FileIngestionMethod::Recursive:
+    case FileSerialisationMethod::Recursive:
         restorePath(path, source);
         break;
     }
@@ -63,7 +102,7 @@ void restorePath(
 
 HashResult hashPath(
     SourceAccessor & accessor, const CanonPath & path,
-    FileIngestionMethod method, HashAlgorithm ha,
+    FileSerialisationMethod method, HashAlgorithm ha,
     PathFilter & filter)
 {
     HashSink sink { ha };
@@ -71,4 +110,20 @@ HashResult hashPath(
     return sink.finish();
 }
 
+
+Hash hashPath(
+    SourceAccessor & accessor, const CanonPath & path,
+    FileIngestionMethod method, HashAlgorithm ht,
+    PathFilter & filter)
+{
+    switch (method) {
+    case FileIngestionMethod::Flat:
+    case FileIngestionMethod::Recursive:
+        return hashPath(accessor, path, (FileSerialisationMethod) method, ht, filter).first;
+    case FileIngestionMethod::Git:
+        return git::dumpHash(ht, accessor, path, filter).hash;
+    }
+
+}
+
 }

src/libutil/file-content-address.hh

@@ -8,37 +8,38 @@
 namespace nix {
 
 /**
- * An enumeration of the main ways we can serialize file system
+ * An enumeration of the ways we can serialize file system
  * objects.
  */
-enum struct FileIngestionMethod : uint8_t {
+enum struct FileSerialisationMethod : uint8_t {
     /**
-     * Flat-file hashing. Directly ingest the contents of a single file
+     * Flat-file. The contents of a single file exactly.
      */
-    Flat = 0,
+    Flat,
+
     /**
-     * Recursive (or NAR) hashing. Serializes the file-system object in
-     * Nix Archive format and ingest that.
+     * Nix Archive. Serializes the file-system object in
+     * Nix Archive format.
      */
-    Recursive = 1,
+    Recursive,
 };
 
 /**
- * Parse a `FileIngestionMethod` by name. Choice of:
+ * Parse a `FileSerialisationMethod` by name. Choice of:
  *
- *  - `flat`: `FileIngestionMethod::Flat`
- *  - `nar`: `FileIngestionMethod::Recursive`
+ *  - `flat`: `FileSerialisationMethod::Flat`
+ *  - `nar`: `FileSerialisationMethod::Recursive`
  *
- * Oppostite of `renderFileIngestionMethod`.
+ * Opposite of `renderFileSerialisationMethod`.
  */
-FileIngestionMethod parseFileIngestionMethod(std::string_view input);
+FileSerialisationMethod parseFileSerialisationMethod(std::string_view input);
 
 /**
- * Render a `FileIngestionMethod` by name.
+ * Render a `FileSerialisationMethod` by name.
  *
- * Oppostite of `parseFileIngestionMethod`.
+ * Opposite of `parseFileSerialisationMethod`.
  */
-std::string_view renderFileIngestionMethod(FileIngestionMethod method);
+std::string_view renderFileSerialisationMethod(FileSerialisationMethod method);
 
 /**
  * Dump a serialization of the given file system object.
@@ -46,26 +47,97 @@ std::string_view renderFileIngestionMethod(FileIngestionMethod method);
 void dumpPath(
     SourceAccessor & accessor, const CanonPath & path,
     Sink & sink,
-    FileIngestionMethod method,
+    FileSerialisationMethod method,
     PathFilter & filter = defaultPathFilter);
 
 /**
- * Restore a serialization of the given file system object.
+ * Restore a serialisation of the given file system object.
  *
  * @TODO use an arbitrary `FileSystemObjectSink`.
  */
 void restorePath(
     const Path & path,
     Source & source,
-    FileIngestionMethod method);
+    FileSerialisationMethod method);
+
 
 /**
  * Compute the hash of the given file system object according to the
  * given method.
  *
- * The hash is defined as (essentially) hashString(ha, dumpPath(path)).
+ * the hash is defined as (in pseudocode):
+ *
+ * ```
+ * hashString(ha, dumpPath(...))
+ * ```
  */
 HashResult hashPath(
+    SourceAccessor & accessor, const CanonPath & path,
+    FileSerialisationMethod method, HashAlgorithm ha,
+    PathFilter & filter = defaultPathFilter);
+
+/**
+ * An enumeration of the ways we can ingest file system
+ * objects, producing a hash or digest.
+ */
+enum struct FileIngestionMethod : uint8_t {
+    /**
+     * Hash `FileSerialisationMethod::Flat` serialisation.
+     */
+    Flat,
+
+    /**
+     * Hash `FileSerialisationMethod::Git` serialisation.
+     */
+    Recursive,
+
+    /**
+     * Git hashing. In particular files are hashed as git "blobs", and
+     * directories are hashed as git "trees".
+     *
+     * Unlike `Flat` and `Recursive`, this is not a hash of a single
+     * serialisation but a [Merkle
+     * DAG](https://en.wikipedia.org/wiki/Merkle_tree) of multiple
+     * rounds of serialisation and hashing.
+     *
+     * @note Git's data model is slightly different, in that a plain
+     * file doesn't have an executable bit, directory entries do
+     * instead. We decide treat a bare file as non-executable by fiat,
+     * as we do with `FileIngestionMethod::Flat` which also lacks this
+     * information. Thus, Git can encode some but all of Nix's "File
+     * System Objects", and this sort of hashing is likewise partial.
+     */
+    Git,
+};
+
+/**
+ * Parse a `FileIngestionMethod` by name. Choice of:
+ *
+ *  - `flat`: `FileIngestionMethod::Flat`
+ *  - `nar`: `FileIngestionMethod::Recursive`
+ *  - `git`: `FileIngestionMethod::Git`
+ *
+ * Opposite of `renderFileIngestionMethod`.
+ */
+FileIngestionMethod parseFileIngestionMethod(std::string_view input);
+
+/**
+ * Render a `FileIngestionMethod` by name.
+ *
+ * Opposite of `parseFileIngestionMethod`.
+ */
+std::string_view renderFileIngestionMethod(FileIngestionMethod method);
+
+/**
+ * Compute the hash of the given file system object according to the
+ * given method.
+ *
+ * Unlike the other `hashPath`, this works on an arbitrary
+ * `FileIngestionMethod` instead of `FileSerialisationMethod`, but
+ * doesn't return the size as this is this is not a both simple and
+ * useful defined for a merkle format.
+ */
+Hash hashPath(
     SourceAccessor & accessor, const CanonPath & path,
     FileIngestionMethod method, HashAlgorithm ha,
     PathFilter & filter = defaultPathFilter);