Improve binary detection

libgit2 optimizes around not loading the content, when there's no
content callbacks configured.

So we aren't able to get a clear answer from radicle-surf if a file in a
diff is a binary or not.

A simple way without doing additional API requests is to check if the
file type is a executable and if there are no hunks (since we don't get
any deltas in binaries), this is a strong
indicator for a binary.

Signed-off-by: Sebastian Martinez <me@sebastinez.dev>

httpd-client/index.ts

@@ -12,6 +12,7 @@ import type {
   CommitHeader,
   Diff,
   DiffContent,
+  DiffFile,
   HunkLine,
 } from "./lib/project/commit.js";
 import type { Issue, IssueState } from "./lib/project/issue.js";
@@ -44,6 +45,7 @@ export type {
   CommitHeader,
   Diff,
   DiffContent,
+  DiffFile,
   DiffResponse,
   HunkLine,
   Issue,

httpd-client/lib/project/commit.ts

@@ -1,5 +1,14 @@
 import type { z } from "zod";
-export type { Commits, HunkLine, Commit, Diff, CommitHeader, DiffContent };
+export type {
+  Commits,
+  HunkLine,
+  Hunks,
+  Commit,
+  Diff,
+  CommitHeader,
+  DiffFile,
+  DiffContent,
+};
 
 import { array, literal, number, object, optional, string, union } from "zod";
 export { commitHeaderSchema, diffSchema, commitSchema, commitsSchema };
@@ -36,6 +45,8 @@ const deletionHunkLineSchema = object({
   type: literal("deletion"),
 });
 
+type DiffFile = z.infer<typeof diffFileSchema>;
+
 const diffFileSchema = object({
   oid: string(),
   mode: union([
@@ -64,6 +75,8 @@ const hunkLineSchema = union([
   contextHunkLineSchema,
 ]);
 
+type Hunks = z.infer<typeof changesetHunkSchema>;
+
 const changesetHunkSchema = object({
   header: string(),
   lines: array(hunkLineSchema),

src/views/projects/Changeset.svelte

@@ -1,5 +1,5 @@
 <script lang="ts">
-  import type { BaseUrl, Diff } from "@httpd-client";
+  import type { BaseUrl, Diff, DiffContent, DiffFile } from "@httpd-client";
 
   import { pluralize } from "@app/lib/pluralize";
 
@@ -39,6 +39,15 @@
     }
     return s.join(", ");
   };
+
+  // Since libgit2 doesn't check binaries thoroughly in diffs,
+  // and we don't want to loop over all diffs in radicle-surf to get a correct answer.
+  // We assume that a `blobExecutable` file with no hunks is a binary file.
+  function getFileType(diff: DiffContent, file: DiffFile): DiffContent["type"] {
+    return file.mode === "blobExecutable" && diff.hunks.length === 0
+      ? "binary"
+      : diff.type;
+  }
 </script>
 
 <style>
@@ -74,7 +83,7 @@
       {baseUrl}
       {revision}
       filePath={file.path}
-      fileDiff={file.diff}
+      fileDiff={{ ...file.diff, type: getFileType(file.diff, file.new) }}
       headerBadgeCaption="added" />
   {/each}
   {#each diff.deleted as file}
@@ -83,7 +92,7 @@
       {baseUrl}
       {revision}
       filePath={file.path}
-      fileDiff={file.diff}
+      fileDiff={{ ...file.diff, type: getFileType(file.diff, file.old) }}
       headerBadgeCaption="deleted" />
   {/each}
   {#each diff.modified as file}
@@ -92,7 +101,7 @@
       {baseUrl}
       {revision}
       filePath={file.path}
-      fileDiff={file.diff} />
+      fileDiff={{ ...file.diff, type: getFileType(file.diff, file.new) }} />
   {/each}
   {#each diff.moved as file}
     {#if file.diff}