This is an automated email from the ASF dual-hosted git repository.
asf-gitbox-commits pushed a commit to branch asf-site
in repository https://gitbox.apache.org/repos/asf/hop-website.git
The following commit(s) were added to refs/heads/asf-site by this push:
new c9cd4b7196 Website updated to 2ff35d7
c9cd4b7196 is described below
commit c9cd4b71961947279214e3cd48ba72d470afe983
Author: jenkins <[email protected]>
AuthorDate: Fri Oct 30 00:04:49 2026 +0000
Website updated to 2ff35d7
---
manual/next/pipeline/spark/paths-and-filesystems.html | 5 ++++-
1 file changed, 4 insertions(+), 1 deletion(-)
diff --git a/manual/next/pipeline/spark/paths-and-filesystems.html
b/manual/next/pipeline/spark/paths-and-filesystems.html
index 3f4c08463f..a80e5dba86 100644
--- a/manual/next/pipeline/spark/paths-and-filesystems.html
+++ b/manual/next/pipeline/spark/paths-and-filesystems.html
@@ -3074,7 +3074,10 @@ Spark never calls <code>HopVfs</code> for Dataset
<code>load</code> / <code>save
<p>On Windows, Hadoop’s local file system sets POSIX permissions by
running <code>winutils.exe</code> from <code>HADOOP_HOME</code> (or the
<code>hadoop.home.dir</code> system property). Spark file output hits that call
when the commit creates the output directory. With neither variable set, the
write fails with <code>HADOOP_HOME and hadoop.home.dir are unset</code>.</p>
</div>
<div class="paragraph">
-<p>When Hop creates the Spark session on Windows and <code>winutils.exe</code>
is not installed, local <code><a href="file:///"
class="bare">file:///</a></code> reads and writes skip those permission
changes. The files keep the permissions of the Hop process. You do not need a
Hadoop installation for <code>local[*]</code>.</p>
+<p>When Hop creates the Spark session on Windows and <code>winutils.exe</code>
is not installed, local <code><a href="file:///"
class="bare">file:///</a></code> reads and writes skip those permission
changes. Listing a directory during the output commit uses Java file checks as
well. Hadoop’s own listing calls <code>NativeIO.Windows.access</code>,
which needs <code>hadoop.dll</code> and throws
<code>UnsatisfiedLinkError</code> when that library is not installed. The files
keep the [...]
+</div>
+<div class="paragraph">
+<p>A <code><a href="file:///" class="bare">file:///</a></code> path is this
same local file system. A variable such as <code>${PROJECT_HOME}</code> is
often stored as a <code><a href="file:///" class="bare">file:///</a></code>
URI, and resolving it keeps that prefix. Spark File Output passes the resolved
URI to Hadoop. A drive path such as <code>C:/output</code> uses the same file
system.</p>
</div>
<div class="paragraph">
<p>If <code>HADOOP_HOME</code> points at a directory that contains
<code>bin/winutils.exe</code>, Hop keeps Hadoop’s own local file system
and permissions are applied. To choose a file system yourself, set
<code>spark.hadoop.fs.file.impl</code> in the run configuration. That entry
wins.</p>
