From d95162259c94412df041482c27f951b038523b05 Mon Sep 17 00:00:00 2001
From: Garvin Hicking <blog@garv.in>
Date: Thu, 16 Nov 2023 12:24:27 +0100
Subject: [PATCH] [BUGFIX] Check Document EntryRoot respecting case sensitivity
 on macOS (#677)

* [BUGFIX] Check Document entry root by respecting case sensitivity on macOS

On macOS filesystems, a file-check against "index.rst"
using $filesystem->has() would return TRUE, when in fact
a file might be stored as "Index.rst". Thus, at this point
we fetch the whole directory list and compare the contents
with if the INDEX_FILE_NAMES entry matches. This ensures
that we get the file with exactly the casing that is returned
from the filesystem.

Background:

Given a file called "Index.rst" exists on macOS, executing this script:

```
<?php
echo (file_exists('Index.rst') ? 'Index.rst exists!' : 'Index.rst missing.') . "\n";
echo (file_exists('index.rst') ? 'index.rst exists!' : 'index.rst missing') . "\n";
```

will yield:

```
Index.rst exists!
index.rst exists!
```

but in fact, "index.rst" does not exist. macOS does this by default to
prevent case sensitivity issues on the Application-level. In macOS,
if a "index.rst" exists, another file called "Index.rst" may NOT exist.

Sadly, also PHP's realpath() does not yield the "true" filename of a saved
file, so there is no way to deduce the real casing of a file, unless
the directory contents are read.

I made a performance benchmark to test the implications of this change.

Funnily, it does not affect speed in a negative way - in fact it sped
up the process for me. I created 27.000 directories with one file
in them, one directory with 16.000 files and one with 6.000 files.

Average runtime on 5 executions on a macBook M1, no other tasks running:

* Without patch (baseline): 419.67s user 10.27s system 99% cpu 7:13.64 total
* With path:                419.38s user 11.07s system 98% cpu 7:14.98 total

So the impact seems to be within usual fault tolerance.

Resolves: https://github.com/TYPO3-Documentation/render-guides/issues/53

* [CLEANUP] Fix code style
---
 .../src/Handlers/ParseDirectoryHandler.php      | 17 +++++++++++++++--
 1 file changed, 15 insertions(+), 2 deletions(-)

diff --git a/packages/guides/src/Handlers/ParseDirectoryHandler.php b/packages/guides/src/Handlers/ParseDirectoryHandler.php
index 281b1a94a..fe4f796b1 100644
--- a/packages/guides/src/Handlers/ParseDirectoryHandler.php
+++ b/packages/guides/src/Handlers/ParseDirectoryHandler.php
@@ -52,10 +52,23 @@ private function getDirectoryIndexFile(
         string $sourceFormat,
     ): string {
         $extension = $sourceFormat;
-        $hasIndexFile = false;
+
+        // On macOS filesystems, a file-check against "index.rst"
+        // using $filesystem->has() would return TRUE, when in fact
+        // a file might be stored as "Index.rst". Thus, at this point
+        // we fetch the whole directory list and compare the contents
+        // with if the INDEX_FILE_NAMES entry matches. This ensures
+        // that we get the file with exactly the casing that is returned
+        // from the filesystem.
+        $contentFromFilesystem = $filesystem->listContents($directory);
+        $hashedContentFromFilesystem = [];
+        foreach ($contentFromFilesystem as $itemFromFilesystem) {
+            $hashedContentFromFilesystem[$itemFromFilesystem['basename']] = true;
+        }
+
         foreach (self::INDEX_FILE_NAMES as $indexName) {
             $indexFilename = sprintf('%s.%s', $indexName, $extension);
-            if ($filesystem->has($directory . '/' . $indexFilename)) {
+            if (isset($hashedContentFromFilesystem[$indexFilename])) {
                 return $indexName;
             }
         }