Undo write if it fails. #15
Conversation
Fixes #14.
Current coverage is
|
| ssize = struct.pack(self.SIZE_FORMAT, len(string)) | ||
| self.f.write(ssize) | ||
| self.f.flush() | ||
| except: |
There was a problem hiding this comment.
perhaps we should catch only IOError here?
|
@ArturGaspar do you think you could come up with a test for this? |
|
@eliasdorneles A proper test would need a mock file that mimics the buffer mechanism and has a size limit. I'll try. |
|
@dangra would love to hear your feelings about this one. :) |
|
+1 I ❤️ it but I can have 🍶 only after a test case. |
|
Btw, queuelib is really to close 100% test coverage, would love anyone volunteer to get there. |
|
@shaneaevans commented (trough other channel) that |
|
Would it make sense to make flushing optional? |
|
Or should we just close this for now? |
Fixes #14.
The
flush()is necessary because of the following scenario:It could however become a performance issue. Could be fixed by a file object that keeps a buffer in memory by itself and only calls write in KB-sized blocks.