[arrow-nanoarrow] branch main updated: fix(extensions/nanoarrow_ipc): Don't release input stream automatically on end of input (#168)

[paleolimbot]

This is an automated email from the ASF dual-hosted git repository.

paleolimbot pushed a commit to branch main
in repository https://gitbox.apache.org/repos/asf/arrow-nanoarrow.git


The following commit(s) were added to refs/heads/main by this push:
     new 373744c  fix(extensions/nanoarrow_ipc): Don't release input stream 
automatically on end of input (#168)
373744c is described below

commit 373744c299d4df8a38cfe745724cbfe4f062d9c5
Author: Dewey Dunnington <de...@dunnington.ca>
AuthorDate: Mon Mar 27 14:16:20 2023 -0400

    fix(extensions/nanoarrow_ipc): Don't release input stream automatically on 
end of input (#168)
    
    Before this PR, the input was released from within `get_next()` in the
    `ArrowArrayStream` implementation. This is unnecessary and prevented the
    reader from being useful in the context of streams that have
    non-standard layouts (e.g., if several streams are concatenated
    together, one could continue calling `get_next()` to "try again"). Also,
    there is a much better chance that `get_next()` is going to get called
    on another thread and it's easier to reason about lifecycles if the
    stream gets closed on `release()` instead.
---
 extensions/nanoarrow_ipc/src/nanoarrow/nanoarrow_ipc_reader.c | 10 ++--------
 1 file changed, 2 insertions(+), 8 deletions(-)

diff --git a/extensions/nanoarrow_ipc/src/nanoarrow/nanoarrow_ipc_reader.c 
b/extensions/nanoarrow_ipc/src/nanoarrow/nanoarrow_ipc_reader.c
index 58d7da9..1b2ad87 100644
--- a/extensions/nanoarrow_ipc/src/nanoarrow/nanoarrow_ipc_reader.c
+++ b/extensions/nanoarrow_ipc/src/nanoarrow/nanoarrow_ipc_reader.c
@@ -343,12 +343,6 @@ static int ArrowIpcArrayStreamReaderGetNext(struct 
ArrowArrayStream* stream,
                                             struct ArrowArray* out) {
   struct ArrowIpcArrayStreamReaderPrivate* private_data =
       (struct ArrowIpcArrayStreamReaderPrivate*)stream->private_data;
-  // Check if we are all done
-  if (private_data->input.release == NULL) {
-    out->release = NULL;
-    return NANOARROW_OK;
-  }
-
   private_data->error.message[0] = '\0';
   
NANOARROW_RETURN_NOT_OK(ArrowIpcArrayStreamReaderReadSchemaIfNeeded(private_data));
 
@@ -356,8 +350,8 @@ static int ArrowIpcArrayStreamReaderGetNext(struct 
ArrowArrayStream* stream,
   int result = ArrowIpcArrayStreamReaderNextHeader(
       private_data, NANOARROW_IPC_MESSAGE_TYPE_RECORD_BATCH);
   if (result == ENODATA) {
-    // If the stream is finished, release the input
-    private_data->input.release(&private_data->input);
+    // Stream is finished either because there is no input or because
+    // end of stream bytes were read.
     out->release = NULL;
     return NANOARROW_OK;
   }