Rework FeedableBodyGerator #949
Conversation
|
I think there's a bug in int capacity = buffer.remaining() - 10; // be safe (we'll have to add size, ending, etc.)
int size = Math.min(nextPart.buffer.remaining(), capacity);WDYT? |
|
At the moment the whole thing relies on the outgoing buffer being big enough to hold the boundaries and at least one byte. The "closing" part has the same problem. |
bomgar
force-pushed
the
master
branch
3 times, most recently
from
e2bae09 to
4bf4fb5
Compare
|
I rewrote parts of it. What do you think? |
| queue.remove(); | ||
| } else { | ||
| readBodyPart(buffer, nextPart); | ||
| reads++; |
There was a problem hiding this comment.
Looks wrong to me: read must return the number of bytes read, while this seems to be returning the number of parts read (and still assuming buffer is large enough).
|
I've added some comments. |
| private final static class BodyPart { | ||
|
|
||
| private void move(ByteBuffer destination, ByteBuffer source) { | ||
| while(destination.hasRemaining() && source.hasRemaining()) { |
There was a problem hiding this comment.
Why not put a whole slice?
|
It doesn't have to return the exact amount of bytes. A positive number is enough. The amount is represented in the ByteBuffer. |
|
The code also does not assume the outgoing buffer is large enough. It stops writing as soon as it is full and continues with the next buffer at the exact same position. That's why i initialize the buffers in the BodyPart. Also the loop is fine. The copy method from one buffer to another does exactly the same except it checks the destination buffer is large enough (not what i want). |
- don't depend on outgoing buffer size - fill the outgoing buffer with all available data the read method became really big.
Oh, right. Then, as AHC2 is still in alpha, I think we should change |
Ultimately, yes (ping @jroper). |
Sounds good to me. |
|
Merging, I'll take care of some clean up. Thanks! |
the read method became really big.