Commits · cba713cd1aa6662edb91d8636b8ae21417028679 · dms3 / go-bitswap

05 Dec, 2014 40 commits
- refactor(util) move block generator · cba713cd
  Brian Tiger Chow authored Nov 26, 2014
```
@whyrusleeping @jbenet

Putting the block generator in a util dir until blocks.

Can't put it in util/testutil because the util/testutil/dag-generator
imports blockservice and blockservice uses the generator.

Tough problem. This'll do for now.

License: MIT
Signed-off-by: Brian Tiger Chow <brian@perfmode.com>
```
  cba713cd
- fix issues in merkledag · 42c3c413
  Jeromy authored Nov 24, 2014
  
  42c3c413
- fix(bitswap/testutils) vendor · 8adecf8e
  Brian Tiger Chow authored Nov 23, 2014
```
License: MIT
Signed-off-by: Brian Tiger Chow <brian@perfmode.com>
```
  8adecf8e
- fix(bs/notifications) use SubOnceEach to provide uniqueness guarantee · 6fe3af11
  Brian Tiger Chow authored Nov 23, 2014
```
License: MIT
Signed-off-by: Brian Tiger Chow <brian@perfmode.com>

vendor forked pubsub to get SubOnceEach

License: MIT
Signed-off-by: Brian Tiger Chow <brian@perfmode.com>
```
  6fe3af11
- fix(bitswap) build-breaking compilation errors · 9c59e0f0
  Brian Tiger Chow authored Nov 23, 2014
```
License: MIT
Signed-off-by: Brian Tiger Chow <brian@perfmode.com>
```
  9c59e0f0
- add a test to blockservice to demonstate GetBlocks failure. · f56f5506
  Jeromy authored Nov 23, 2014
  
  f56f5506
- ensure sending of wantlist to friendly peers · fe048093
  Jeromy authored Nov 22, 2014
  
  fe048093
- fix(bs/notifications) prevent duplicates · be976cc5
  Brian Tiger Chow authored Nov 21, 2014
```
@whyrusleeping now notifications _guarantees_ there won't be any
duplicates

License: MIT
Signed-off-by: Brian Tiger Chow <brian@perfmode.com>
```
  be976cc5
- test(bs/n) check for duplicates received · 6a7c1d4d
  Brian Tiger Chow authored Nov 21, 2014
```
License: MIT
Signed-off-by: Brian Tiger Chow <brian@perfmode.com>
```
  6a7c1d4d
- fix(bs/n) remove unnecessary variable · 365044ce
  Brian Tiger Chow authored Nov 21, 2014
```
to remove ambiguity

(before it was possible to loop over either topics or keys

by only keeping keys, there's no confusing about what to use for the
loop range

License: MIT
Signed-off-by: Brian Tiger Chow <brian@perfmode.com>
```
  365044ce
- misc(bs/n) rm dead code · 173ac606
  Brian Tiger Chow authored Nov 21, 2014
```
License: MIT
Signed-off-by: Brian Tiger Chow <brian@perfmode.com>
```
  173ac606
- refactor(bitswap) forwardN no longer needed · a50a3497
  Brian Tiger Chow authored Nov 21, 2014
```
@whyrusleeping

now, the pubsub channel closes after sending N blocks. we got this
functionality for free from the fix. So, the forwardN wrap is no longer
required! woohoo

License: MIT
Signed-off-by: Brian Tiger Chow <brian@perfmode.com>
```
  a50a3497
- tests(bitswap/notifications) test niladic · abdf5c87
  Brian Tiger Chow authored Nov 21, 2014
```
License: MIT
Signed-off-by: Brian Tiger Chow <brian@perfmode.com>
```
  abdf5c87
- fix(bitswap/notifications) subscribe to many · 097bc1b4
  Brian Tiger Chow authored Nov 21, 2014
```
License: MIT
Signed-off-by: Brian Tiger Chow <brian@perfmode.com>
```
  097bc1b4
- docs(bitswap/notifications) · a6371b4b
  Brian Tiger Chow authored Nov 21, 2014
```
License: MIT
Signed-off-by: Brian Tiger Chow <brian@perfmode.com>
```
  a6371b4b
- use @maybebtc's ForwardBlocks function · a2a4327b
  Jeromy authored Nov 21, 2014
  
  a2a4327b
- test(notifications) · f1eb07d9
  Brian Tiger Chow authored Nov 21, 2014
```
we expect this to fail. will be fixed in upcoming commit

License: MIT
Signed-off-by: Brian Tiger Chow <brian@perfmode.com>
```
  f1eb07d9
- a little more correctness on the new bitswap impl · 92e2d247
  Jeromy authored Nov 21, 2014
  
  92e2d247
- tracking down a bug dhthell found, added asserts and better logging. · 36798def
  Jeromy authored Nov 21, 2014
  
  36798def
- wire GetBlocks into blockservice · b5121d63
  Jeromy authored Nov 21, 2014
  
  b5121d63
- test(bitswap) Close (but skip for now) · b90e5fb7
  Brian Tiger Chow authored Nov 20, 2014
```
License: MIT
Signed-off-by: Brian Tiger Chow <brian@perfmode.com>
```
  b90e5fb7
- refactor(bitswap) group the deferreds · 58b9745f
  Brian Tiger Chow authored Nov 20, 2014
```
License: MIT
Signed-off-by: Brian Tiger Chow <brian@perfmode.com>
```
  58b9745f
- fix(exchange) allow exchange to be closed · 26e6f28e
  Brian Tiger Chow authored Nov 20, 2014
```
License: MIT
Signed-off-by: Brian Tiger Chow <brian@perfmode.com>
```
  26e6f28e
- fix(bitswap) signal termination to async'ly spawned workers · d6e2157a
  Brian Tiger Chow authored Nov 20, 2014
```
License: MIT
Signed-off-by: Brian Tiger Chow <brian@perfmode.com>
```
  d6e2157a
- style(bitswap) name -> loop · 5045d096
  Brian Tiger Chow authored Nov 20, 2014
```
eh?

License: MIT
Signed-off-by: Brian Tiger Chow <brian@perfmode.com>
```
  5045d096
- feat(bitswap) loop over all provided keys · 499aa2c3
  Brian Tiger Chow authored Nov 20, 2014
```
License: MIT
Signed-off-by: Brian Tiger Chow <brian@perfmode.com>
```
  499aa2c3
- feat(bitswap) find providers for all keys on wantlist · 36ebde35
  Brian Tiger Chow authored Nov 20, 2014
```
@jbenet @whyrusleeping

this addresses a failure case where

1) bitswap wants blocks A and B
2) partner 1 has A and partner 2 has B
3) We choose a key at random, drawing A.
4) Then, we request A, neglecting to find a provider for B.

Sending the full wantlist is meant to be used as a helpful additional
piece of data, but...

unless our hunch is support by statistical inference at runtime,
it's not safe to assume that a peer will have blocks for related keys.
Routing must be the source of truth.

License: MIT
Signed-off-by: Brian Tiger Chow <brian@perfmode.com>
```
  36ebde35
- fix(bitswap/loop) add to wantlist just once · 6af9f417
  Brian Tiger Chow authored Nov 20, 2014
```
oops

set Add is idempotent but it's a waste of resources

License: MIT
Signed-off-by: Brian Tiger Chow <brian@perfmode.com>
```
  6af9f417
- rename · 7d72132c
  Brian Tiger Chow authored Nov 20, 2014
```
License: MIT
Signed-off-by: Brian Tiger Chow <brian@perfmode.com>
```
  7d72132c
- refactor(blockstore) mv under blocks/ · 08d0a11b
  Brian Tiger Chow authored Nov 20, 2014
```
@jbenet @whyrusleeping

the pyramids were built one brick at a time

addresses: https://github.com/jbenet/go-ipfs/issues/370

License: MIT
Signed-off-by: Brian Tiger Chow <brian@perfmode.com>
```
  08d0a11b
- tests(bitswap) share code between the two large tests · 80244dac
  Brian Tiger Chow authored Nov 20, 2014
```
License: MIT
Signed-off-by: Brian Tiger Chow <brian@perfmode.com>
```
  80244dac
- fix(bitswap) stop the ticker when the run loop exits · fd8e2b3a
  Brian Tiger Chow authored Nov 19, 2014
```
@whyrusleeping

License: MIT
Signed-off-by: Brian Tiger Chow <brian@perfmode.com>
```
  fd8e2b3a
- feat(bitswap) implement GetBlocks · 2a4a6d3e
  Brian Tiger Chow authored Nov 19, 2014
```
@whyrusleeping @jbenet

License: MIT
Signed-off-by: Brian Tiger Chow <brian@perfmode.com>
```
  2a4a6d3e
- refactor(bitswap) move wantlist to loop receive · 470b02d4
  Brian Tiger Chow authored Nov 19, 2014
```
License: MIT
Signed-off-by: Brian Tiger Chow <brian@perfmode.com>
```
  470b02d4
- tests(bitswap) share constructor between tests · 12b83ff8
  Brian Tiger Chow authored Nov 19, 2014
```
@whyrusleeping i hope this makes it a bit easier to work with tests

License: MIT
Signed-off-by: Brian Tiger Chow <brian@perfmode.com>
```
  12b83ff8
- feat(bitswap/notifications) Subscribe to multiple keys · 8a87b709
  Brian Tiger Chow authored Nov 19, 2014
```
License: MIT
Signed-off-by: Brian Tiger Chow <brian@perfmode.com>
```
  8a87b709
- style(bitswap/notifications) make it more obvious · 140a141c
  Brian Tiger Chow authored Nov 19, 2014
```
License: MIT
Signed-off-by: Brian Tiger Chow <brian@perfmode.com>
```
  140a141c
- randomize rebroadcast target · 6ee1fe5e
  Jeromy authored Nov 20, 2014
  
  6ee1fe5e
- remove buffer timing in bitswap in favor of manual batching · 26f78574
  Jeromy authored Nov 20, 2014
  
  26f78574
- fix tests halting · 44f32138
  Jeromy authored Nov 19, 2014
  
  44f32138