Balance assertion tolerances

[Taylor Corum]

Hello,

According to what I have read in the docs (https://beancount.github.io/docs/precision_tolerances.html#balance-assertions-padding), on balance assertions "The tolerance is inferred automatically to be 1 unit of the least significant digit of the number on the balance assertion." The tolerance on regular transactions appears to be half of that.

What is the rational for inferring the tolerance to 1 unit of the least significant digit rather than half that much? This seems a little unintuitive to me and I would like to make sure I understand. For example the following balance assertion would pass even though it is off by one cent.

2021-01-01 * "Initial transaction"
    Assets:Cash              0.49 USD
    Equity:Opening-Balances

2021-01-02 balance Assets:Cash  0.50 USD

Is it best practice to use balance assertions with a higher degree of precision than the base currency or to explicitly give a tolerance on each assertion?

Thank you for your help.

[Nick Reilingh]

Same question here -- I take it this makes more sense for investment accounts and atypical currencies, but I don't see how it makes sense for (e.g.) USD when asserting conformance to a bank account statement. Further nonintuitive behavior is that if you enter a balance assertion at a whole number and don't add any decimal places, it is automatically set to zero tolerance, that is:

2021-01-01 * "Initial transaction"
    Assets:Cash              0.49 USD

    Income:Other

2021-01-02 balance Assets:Cash  0.50 USD ; passes

2021-01-02 * "second transaction"
    Assets:Cash              0.50 USD
    Income:Other

2021-01-03 balance Assets:Cash  1 USD ; fails

I think I will probably be inputting/importing my future bank statement balance assertions with explicit zero tolerance (0.50~0 USD).

[Martin Blais]

It was designed that way because when using no digits the typical use case is for counting discrete things and a zero tolerance is what's needed. Beancount doesn't know that "USD" is anything more special than "WIDGETS."  Possible fixes:

- changing the rule so that inference of zero tolerance only occurs if there are never any other amounts for that currency in the file with some sub unit precision

- finally implementing the explicit precision and tolerance options (I started a few months ago, didn't finish, too busy)

--
You received this message because you are subscribed to the Google Groups "Beancount" group.
To unsubscribe from this group and stop receiving emails from it, send an email to beancount+...@googlegroups.com.
To view this discussion visit https://groups.google.com/d/msgid/beancount/c2506461-29b2-405f-80d4-4a69cfbfac74n%40googlegroups.com.