You should be able to run this engine like

cargo run -- samples/transactions.csv > samples/test.output.csv 

The input file is the only argument. Output is std out.

To enable logger use

RUST_LOG=<logger-option> cargo run -- samples/transactions.csv > samples/test.output.csv

<logger-option>: error, info and debug

To run tests use

RUST_LOG=debug cargo test -- --nocapture

Operations:

• deposit - regular deposit transaction
• withdraw - regular withdraw transaction
• dispute - user open dispute
• resolve - resolve dispute releasing the founds helded and increasing available balance
• chargeback - resolve dispute reversing the transaction (withdraw or deposit) and block user to check fraud

Separate transactions by accounts, to avoid unnecessary costs in search operations. We have 3 different operations that do searches: dispute, solve and reverse.

Spawn threads to handle the summary function quickly (use subsets of accounts). Done

Try different approach to read CSV's

Library Documentation

Integration tests

In case you receive many requests to this program, you must use an SQS or similar to handle the process in the background, for greater efficiency you could use a file provider to save ".csv" files (S3) and use a Worker/Consumer to read from the SQS and get these files from their location.

In my opinion chargebacks shouldn't be processed in batch files, to avoid fraud maybe it should use a warning in the logger and include the chargeback transactions in another file to be processed after careful verification.