beancount-import VS beangulp

Fighting with Mint and categorizing Amazon purchases was what initially pushed me down the path into plain text accounting (PTA).

I ended up long down the rabbit hole with auto-downloading Amazon orders (originally with https://github.com/jbms/finance-dl, but then my own custom scraping) and importing and matching them up with credit card transactions using beancount-import (https://github.com/jbms/beancount-import).

This ultimately resulted in me spending a lot less on Amazon - to the point that now doing it manually wouldn't be too bad...

I personally started with beancount-import to process QFX/OFX files from my bank. It's a nice product, because it automatically joins both sides of a transaction for transfers/payments from one bank to another, and has some auto-categorization features. But it does take some work to understand and get running.

Some skimming of the OFX import code indicates that the importer manages the duplicate detection itself. Since it looks like you need to pretty much write a CSV importer yourself, you'd probably have to manage the uniqueness using some identifier from the CSV file.

These HTML are stored in a common folder, then parsed by beancount-import using the built-in importer for Amazon Invoice HTML files from that project. The author of that project wrote/uses finance-dl to get his Amazon invoices. At one point, his download code wasn't working for me, so I wrote my own. Independently, for reasons I no longer remember.

Now, I'm trying to set up the bean-import (https://github.com/jbms/beancount-import) module but i have run in to some problems.

I handle them with Beancount and beancount-import. beancount-import identifies duplicate transactions and merges them automatically. It handles both the case where two import sources have the same transaction (e.g. your bank account/credit card situation) and the case where there's an existing manually-entered transaction that also exists in an import source.

What I use a lot is https://github.com/jbms/beancount-import. Since I know a bit of python I was able tweak the importer for my bank statements. It helps a lot.

I use beancount-import which does a great job of allowing for multiple sources of data (including manual entries) and matching up postings from multiple sources. It helps a lot for my particular use.

It's the wrong underlying toolset (beancount instead of ledger), but otherwise you're describing how beancount-import works.

I use beangulp, the importer framework for beancount, and then convert to ledger with beancount2ledger.

The main goals for Beancount v3 are rewriting some core parts in C++ to improve performance, an improved parser, and fixing assorted issue that require breaking backward compatibility, splitting some parts of Beancount into separate packages for better decoupling of functionalities and ease of maintenance. The work on the C++ code and the parser has started but that is not hooked up to the main Python entry points yet, thus there is not risk of regressions due to that. For switching to v3 the main change ist that some code has been split out to separate projects: most notably the ingest framework has become beangulp https://github.com/beancount/beangulp.

But it appears the example links are broken. They may have moved here, in prep for beancount v3: https://github.com/beancount/beangulp/tree/master/examples