Batch Validation Missed Addenda Format

[hshih-lead]

ACH Version

1.45.5

What were you trying to do?

We rely on the batch validation to make sure we're structuring our NACHA file with the correct format, using this validation call:
https://github.com/moov-io/ach/blob/master/batcher.go#L46

What did you expect to see?

We expected that call to validate the batch format catching all errors. More specifically, the addendaRecordIndicator and the number of addendas in the entry should match.

What did you see?

We accidentally set the addendaRecordIndicator field as 1 in the entry detail without actually having addendas set. This bad format led to entries being rejected by the Fed via return code R25.

How can we reproduce the problem?

Set the addendaRecordIndicator to a value that does not match the actual number of addendas in the entry. The batch validation will not catch that as an error.

[adamdecaf]

What was the SEC code of the batch? Any/All of them?

Was this using the HTTP api (so JSON -> *ach.File) or the Reader type on a file?

[hshih-lead]

What was the SEC code of the batch? Any/All of them?

For our particular incident, it happened on WEB, PPD, and CCD. But really it could be any SEC code.

Was this using the HTTP api (so JSON -> *ach.File) or the Reader type on a file?

It's for both, as we have Reader type that came from file reading as well as JSON -> File use case

[adamdecaf]

Do you have a sample file that recreates this? I'll try and create one.

[adamdecaf]

Implementing this validation is tricky.. In the reader we have to delay validation until the addenda records have been read, but in the writer validation needs to occur before we write the entry record.. There's not a great way to implement this inside of Validate() as-is today.

[adamdecaf]

I haven't been able to figure out a decent way to validate AddendaRecordIndicator only when it's possible to validate it. Inside the reader/Validate() isn't quite good enough. We may need some private validation functions that are called at the exact right time.

You can try out what I've gotten so far and improve on it #1655

[hshih-lead]

I haven't been able to figure out a decent way to validate AddendaRecordIndicator only when it's possible to validate it. Inside the reader/Validate() isn't quite good enough. We may need some private validation functions that are called at the exact right time.

You can try out what I've gotten so far and improve on it #1655

Makes sense to have a separate validation function, given the constraint of the existing Validate() function. Will give it a try and see what are some good ways to handle this!

[adamdecaf]

A separate function that's private we can call. I don't want to expose multiple validation methods to the public.

[takhleeq-ai]

I’d like to take this; will open a PR shortly.

[takhleeq-ai]

I’d like to take this; will open a PR shortly.

PR merge requested. Awaiting review.