Unchecked cleanup #3600
Unchecked cleanup #3600
Conversation
…d_ptr<nano::block> upcast. This allows it to be implicitly convertible to nano::unchecked_info through a conversion-constructor.
Since the hash in unchecked_key must always match the block hash of the block in unchecked_info, put don't directly accept unchecked_key values and instead infers it from the block in unchecked_info directly. Only the dependency needs to be passed in.
…ally ordered instead of totally ordered which doesn't match lmdb behavior. It's not established this has a performance impact and this table is a candidate for using a disk based hash table for superior performance as it doesn't need atomic guarantees.
…unction nano::unchecked_store::equal_range which returns a pair of iterators covering keys equal to the given dependency.
… iterators covering all items in the unchecked store.
| @@ -352,7 +352,7 @@ TEST (bootstrap, simple) | |||
| nano::logger_mt logger; | |||
| auto store = nano::make_store (logger, nano::unique_path (), nano::dev::constants); | |||
| ASSERT_TRUE (!store->init_error ()); | |||
| auto block1 (std::make_shared<nano::send_block> (0, 1, 2, nano::keypair ().prv, 4, 5)); | |||
| std::shared_ptr<nano::block> block1 = std::make_shared<nano::send_block> (0, 1, 2, nano::keypair ().prv, 4, 5); | |||
There was a problem hiding this comment.
This clean up looks good, but since you touched the lines anyway, why not auto block = std::make_shared...? Not really important but just in the spirit of if cleaning it up, clean it all the way :D.
There was a problem hiding this comment.
This one specifically is because std::shared_ptrnano::send_block won't implicitly convert to unchecked_info when passed to unchecked_store::put. It won't do an upcast and also an implicit conversion.
| send5->signature.bytes[31] ^= 0x1; | ||
| auto signature = send5->block_signature (); | ||
| signature.bytes[31] ^= 0x1; | ||
| send5->signature_set (signature); |
There was a problem hiding this comment.
I agree it's more clean to get the signature, touch it, then set it back, but was there a functional problem with changing it in-place?
There was a problem hiding this comment.
Since we're holding a pointer to nano::block instead of the specific block type, the signature field is no longer directly accessible.
|
Looks good to me; also, reviewing commits individually turned out to be really helpful, so I recommend that other reviewers do that. |
This is a series of commits in preparation for adding the unchecked_map ADT.
Each commit is best reviewed individually.