Re: [PR] fix(arrow/ipc): make payload write failures terminal [arrow-go]

["zeroshade (via GitHub)" <gi...@apache.org>]

zeroshade commented on code in PR #1047:
URL: https://github.com/apache/arrow-go/pull/1047#discussion_r3722058653


##########
arrow/ipc/writer.go:
##########
@@ -148,24 +165,38 @@ func (w *Writer) Close() error {
 	}
 
 	err := w.pw.Close()
+	w.pw = nil
+	w.releaseDictionaries()
 	if err != nil {
-		return fmt.Errorf("arrow/ipc: could not close payload writer: %w", err)
+		return w.fail(fmt.Errorf("arrow/ipc: could not close payload writer: %w", err))
 	}
-	w.pw = nil
 
+	return nil
+}
+
+func (w *Writer) releaseDictionaries() {
 	for _, d := range w.lastWrittenDicts {
 		d.Release()
 	}
+	w.lastWrittenDicts = nil
+}
 
-	return nil
+func (w *Writer) fail(err error) error {
+	if w.err == nil {
+		w.err = err
+	}
+	return w.err

Review Comment:
   ```suggestion
       w.err = errors.Join(w.err, err)
   	return w.err
   ```
   
   otherwise we can lose the error if we end up with multiple errors



##########
arrow/ipc/writer.go:
##########
@@ -291,10 +330,11 @@ func (w *Writer) start() error {
 	for _, data := range ps {
 		err := w.pw.WritePayload(data)
 		if err != nil {
-			return err
+			return w.fail(err)
 		}
 	}
 
+	w.started = true

Review Comment:
   we should probably move `w.started = true` above the for loop here since if even one succeeds writing then we technically started. The way this is currently written, if we succeed on one payload, but then fail on a subsequent payload `w.started` will remain `false`.



##########
arrow/ipc/writer.go:
##########
@@ -59,6 +59,17 @@ func (w *streamWriter) WritePayload(p Payload) error {
 func (w *streamWriter) Write(p []byte) (int, error) {
 	n, err := w.w.Write(p)
 	w.pos += int64(n)
+	if err == nil && n != len(p) {
+		err = io.ErrShortWrite
+	}
+	return n, err
+}
+
+func writeFull(w io.Writer, p []byte) (int, error) {
+	n, err := w.Write(p)
+	if err == nil && n != len(p) {
+		err = io.ErrShortWrite
+	}
 	return n, err
 }

Review Comment:
   we ignore the n on every call to `writeFull`, should we bother even returning n here?



-- 
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.

To unsubscribe, e-mail: github-unsubscribe@arrow.apache.org

For queries about this service, please contact Infrastructure at:
users@infra.apache.org