1 # Application BlockChain Interface (ABCI)
3 [![CircleCI](https://circleci.com/gh/tendermint/abci.svg?style=svg)](https://circleci.com/gh/tendermint/abci)
5 Blockchains are a system for multi-master state machine replication.
6 **ABCI** is an interface that defines the boundary between the replication engine (the blockchain),
7 and the state machine (the application).
8 By using a socket protocol, we enable a consensus engine running in one process
9 to manage an application state running in another.
14 go get github.com/tendermint/abci
15 cd $GOPATH/src/github.com/tendermint/abci
20 For background information on ABCI, motivations, and tendermint, please visit [the documentation](http://tendermint.readthedocs.io/en/master/).
21 The two guides to focus on are the `Application Development Guide` and `Using ABCI-CLI`.
23 Previously, the ABCI was referred to as TMSP.
25 The community has provided a number of addtional implementations, see the `Tendermint Ecosystem` in [the documentation](http://tendermint.readthedocs.io/en/master/).
29 We provide three implementations of the ABCI in Go:
37 ABCI is best implemented as a streaming protocol.
38 The socket implementation provides for asynchronous, ordered message passing over unix or tcp.
39 Messages are serialized using Protobuf3 and length-prefixed.
40 Protobuf3 doesn't have an official length-prefix standard, so we use our own. The first byte represents the length of the big-endian encoded length.
42 For example, if the Protobuf3 encoded ABCI message is `0xDEADBEEF` (4 bytes), the length-prefixed message is `0x0104DEADBEEF`. If the Protobuf3 encoded ABCI message is 65535 bytes long, the length-prefixed message would be like `0x02FFFF...`.
46 GRPC is an rpc framework native to Protocol Buffers with support in many languages.
47 Implementing the ABCI using GRPC can allow for faster prototyping, but is expected to be much slower than
48 the ordered, asynchronous socket protocol.
50 Note the length-prefixing used in the socket implementation does not apply for GRPC.
54 The simplest implementation just uses function calls within Go.
55 This means ABCI applications written in Golang can be compiled with TendermintCore and run as a single binary.
59 The `abci-cli` tool wraps any ABCI client and can be used for probing/testing an ABCI application.
60 See [the documentation](http://tendermint.readthedocs.io/en/master/) for more details.
62 Multiple example apps are included:
63 - the `abci-cli counter` application, which illustrates nonce checking in txs
64 - the `abci-cli dummy` application, which illustrates a simple key-value merkle tree
65 - the `abci-cli dummy --persistent` application, which augments the dummy with persistence and validator set changes
69 The [primary specification](https://github.com/tendermint/abci/blob/master/types/types.proto) is made using Protocol Buffers.
76 See `protoc --help` and [the Protocol Buffers site](https://developers.google.com/protocol-buffers/) for details on compiling for other languages.
77 Note we also include a [GRPC](http://www.grpc.io/docs) service definition.
79 For the specification as an interface in Go, see the [types/application.go file](https://github.com/tendermint/abci/blob/master/types/application.go).
83 ABCI requests/responses are defined as simple Protobuf messages in [this schema file](https://github.com/tendermint/abci/blob/master/types/types.proto).
84 TendermintCore sends the requests, and the ABCI application sends the responses.
85 Here, we describe the requests and responses as function arguments and return values, and make some notes about usage:
89 * `Data ([]byte)`: The request transaction bytes
91 * `Code (uint32)`: Response code
92 * `Data ([]byte)`: Result bytes, if any
93 * `Log (string)`: Debug or error message
95 Append and run a transaction. If the transaction is valid, returns CodeType.OK
99 * `Data ([]byte)`: The request transaction bytes
101 * `Code (uint32)`: Response code
102 * `Data ([]byte)`: Result bytes, if any
103 * `Log (string)`: Debug or error message
105 Validate a mempool transaction, prior to broadcasting or proposing. This message should not mutate the main state, but application
106 developers may want to keep a separate CheckTx state that gets reset upon Commit.
108 CheckTx can happen interspersed with DeliverTx, but they happen on different ABCI connections - CheckTx from the mempool connection, and DeliverTx from the consensus connection. During Commit, the mempool is locked, so you can reset the mempool state to the latest state after running all those DeliverTxs, and then the mempool will re-run whatever txs it has against that latest mempool state.
110 Transactions are first run through CheckTx before broadcast to peers in the mempool layer.
111 You can make CheckTx semi-stateful and clear the state upon `Commit` or `BeginBlock`,
112 to allow for dependent sequences of transactions in the same block.
116 * `Data ([]byte)`: The Merkle root hash
117 * `Log (string)`: Debug or error message
119 Return a Merkle root hash of the application state.
123 * `Data ([]byte)`: Raw query bytes. Can be used with or in lieu of Path.
124 * `Path (string)`: Path of request, like an HTTP GET path. Can be used with or in liue of Data.
125 * Apps MUST interpret '/store' as a query by key on the underlying store. The key SHOULD be specified in the Data field.
126 * Apps SHOULD allow queries over specific types like '/accounts/...' or '/votes/...'
127 * `Height (uint64)`: The block height for which you want the query (default=0 returns data for the latest committed block). Note that this is the height of the block containing the application's Merkle root hash, which represents the state as it was after committing the block at Height-1
128 * `Prove (bool)`: Return Merkle proof with response if possible
130 * `Code (uint32)`: Response code
131 * `Key ([]byte)`: The key of the matching data
132 * `Value ([]byte)`: The value of the matching data
133 * `Proof ([]byte)`: Proof for the data, if requested
134 * `Height (uint64)`: The block height from which data was derived. Note that this is the height of the block containing the application's Merkle root hash, which represents the state as it was after committing the block at Height-1
135 * `Log (string)`: Debug or error message
136 *Please note* The current implementation of go-merkle doesn't support querying proofs from past blocks, so for the present moment, any height other than 0 will return an error (recall height=0 defaults to latest block). Hopefully this will be improved soon(ish)
140 * `Data (string)`: Some arbitrary information
141 * `Version (Version)`: Version information
142 * `LastBlockHeight (uint64)`: Latest block for which the app has called Commit
143 * `LastBlockAppHash ([]byte)`: Latest result of Commit
146 Return information about the application state. Used to sync the app with Tendermint on crash/restart.
150 * `Key (string)`: Key to set
151 * `Value (string)`: Value to set for key
153 * `Log (string)`: Debug or error message
155 Set application options. E.g. Key="mode", Value="mempool" for a mempool connection, or Key="mode", Value="consensus" for a consensus connection.
156 Other options are application specific.
160 * `Validators ([]Validator)`: Initial genesis validators
162 Called once upon genesis
166 * `Hash ([]byte)`: The block's hash. This can be derived from the block header.
167 * `Header (struct{})`: The block header
169 Signals the beginning of a new block. Called prior to any DeliverTxs. The header is expected to at least contain the Height.
173 * `Height (uint64)`: The block height that ended
175 * `Diffs ([]Validator)`: Changed validators with new voting powers (0 to remove)
177 Signals the end of a block. Called prior to each Commit after all transactions. Validator set is updated with the result.
181 * `Message (string)`: A string to echo back
183 * `Message (string)`: The input string
185 * Echo a string to test an abci client/server implementation
189 * Signals that messages queued on the client should be flushed to the server. It is called periodically by the client implementation to ensure asynchronous requests are actually sent, and is called immediately to make a synchronous request, which returns when the Flush response comes back.