How to insert documents to couchdb in bulk by chaincode - hyperledger-fabric

Question

How to insert/write documents to couchdb in bulk by chaincode? It seems that the chaincode shim library(https://godoc.org/github.com/hyperledger/fabric/core/chaincode/shim) doesn't have such API.
For reading documents, it seems that there is an API named "GetQueryResult"(https://godoc.org/github.com/hyperledger/fabric/core/chaincode/shim#ChaincodeStub.GetQueryResult). In the string parameter "query", we can construct bulk get request.
But for inserting/writing documents, is there any bulk API for chaincode? Thanks in advance.

Answer 1

When chaincode executes, each PutState() adds a key/value to the transaction's proposed writeset. You can call PutState() multiple times in a chaincode, and the final set of keys/values will appear in the transaction's proposed writeset. Note that nothing is written to CouchDB at chaincode execution time.
Only when the transaction is submitted for ordering, will the transaction appear in a block that gets processed by all peers. Each peer validates the transactions and then applies the writesets of all valid transactions in the block to the CouchDB state database. Note that the state database commit is indeed using the CouchDB bulk update API (HTTP _bulk_docs), so you automatically get the desired bulk update performance in CouchDB.
If there are many key/value updates in the block, Fabric will actually group them into batches of 1000 (configurable using core.yaml maxBatchUpdateSize property) when committing to CouchDB to avoid any issues with excessively large payloads. Finally, to ensure all writes appear to Fabric as an atomic commit, Fabric writes a final savepoint per block to the CouchDB state database and flushes the CouchDB data to disk. This ensures any chaincode executions get a consistent view of the state data, and ensures Fabric can recover from any peer crash with full data integrity.

Answer 2

If you mean you want to batch multiple transactions in one batch then it is not advisable to do so because if they are updating the same value then while looking at the history, the final change can only be tracked. So, it is better to submit individual transactions.

Answer 3

Not sure if this answers your question but by document, lets assume you mean a PDF file.
This has two functions, read a key value pair and write one. Nothing special here.
'use strict';
const { Contract } = require('fabric-contract-api');
Class SomeName extends Contract {
async initLedger(ctx) { }
async writeDocument(ctx, documentId, documentAsString) {
await ctx.stub.putState(documentId, Buffer.from(documentAsString));
};
async getDocument(ctx, documentId) {
const docAsBytes = await ctx.stub.getState(documentId);
return carAsBytes.toString();
};
}
Now comes the server side code
const { FileSystemWallet, Gateway } = require('fabric-network');
async main() {
const walletPath = path.join(process.cwd(), 'wallet');
const wallet = new FileSystemWallet(walletPath); // can be wherever your path is
const gateway = new Gateway();
await gateway.connect(ccp, { wallet, identity: 'user1', discovery: { enabled: false } });
const network = await gateway.getNetwork('mychannel');
// Get the contract from the network.
const contract = network.getContract('SomeName');
// Now over here lets assume you get the PDF file from a POST request using Multer
// We get the PDF file as a buffer, then convert to a string and send it
let pdfFile = req.file;
await contract.submitTransaction('writeDocument', req.file.buffer.toString('utf-8);
}
and if you want to retrieve the PDF file
const getDocument = async (req, res, next) => {
const walletPath = path.join(process.cwd(), 'wallet');
const wallet = new FileSystemWallet(walletPath); // can be wherever your path is
const gateway = new Gateway();
await gateway.connect(ccp, { wallet, identity: 'user1', discovery: { enabled: false } });
const network = await gateway.getNetwork('mychannel');
// Get the contract from the network.
const contract = network.getContract('SomeName');
const pdfAsString= await contract.evaluateTransaction('getDocument', req.body.documentId);
// Now do whatever with this
};
So I hope this answers your question regarding document upload, that being said its not a good practice to upload documents on a blockchain as it can slow down the transaction speed significantly.
In that case, you can store the document on a local server such as mongoDB and then refer a sha256 hash of the document on the blockchain to cross check the authenticity of the document at a later stage.