Resume partial blob downloads #449
Conversation
| byte buf[] = new byte[4096]; | ||
| // resume unfinished downloads | ||
| File outFile = new File(tempPath.toAbsolutePath().toString()); | ||
| is.skip(outFile.length()); |
There was a problem hiding this comment.
I suppose this is better than we have currently, but I'm guessing this results in effectively the same behavior from a stream consumption perspective, i.e. we have to read the bytes (even if we drop them)?
| @@ -223,12 +226,17 @@ public InputStream getInputStream() throws IOException { | |||
| @Override | |||
| public InputStream getInputStream() throws IOException { | |||
|
|
|||
| Path tempPath = path.resolveSibling(path.getName(path.getNameCount()-1) + "-" + UUID.randomUUID().toString()); | |||
| Path tempPath = path.resolveSibling(PARTIAL_DOWNLOAD + path.getName(path.getNameCount()-1)); | |||
There was a problem hiding this comment.
nice, although FWIW most partial download files I've seen end in .part or similar (rather than being prefixed)
There was a problem hiding this comment.
There is some fragile filename parsing at other places in code, and it seemed to work better to have partial in the prefix as opposed to suffix. One eg. is checking if filename startswith "snapshot.".
| HollowConsumer consumer = consumerBuilder.build(); | ||
| boolean downloadInterrupted = false; | ||
| try { | ||
| consumer.triggerRefreshTo(testVersion); |
There was a problem hiding this comment.
minor cleanup - you could move the fail call here and not need the downloadInterrupted boolean at all
This is useful for local dev on spotty internet connection