Better error messages when `beancount` format does not apply

[tbm]

It's late and maybe I'm doing something wrong but I get a Python syntax error with `.format beancount``:

beanquery> .format beancount
beanquery> SELECT date, META("receipt")  WHERE account ~ "^Expenses:" 

gives me:

  File "/home/tbm/.local/lib/python3.11/site-packages/beanquery/render/beancount.py", line 11, in render
    return printer.print_entries([entry for entry, in rows], dcontext, file=file)
                                 ^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/home/tbm/.local/lib/python3.11/site-packages/beanquery/render/beancount.py", line 11, in <listcomp>
    return printer.print_entries([entry for entry, in rows], dcontext, file=file)
                                            ^^^^^^
ValueError: too many values to unpack (expected 1)

After removing the comma after entry,, I get the following (note that I'm on beancount 2.x):

  File "/home/tbm/.local/lib/python3.11/site-packages/beanquery/shell.py", line 595, in on_Select
    return render(desc, rows, out, dcontext=dcontext, **self.settings.todict())
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/home/tbm/.local/lib/python3.11/site-packages/beanquery/render/beancount.py", line 11, in render
    return printer.print_entries([entry for entry in rows], dcontext, file=file)
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/home/tbm/.local/lib/python3.11/site-packages/beancount/parser/printer.py", line 423, in print_entries
    string = eprinter(entry)
             ^^^^^^^^^^^^^^^
  File "/home/tbm/.local/lib/python3.11/site-packages/beancount/parser/printer.py", line 123, in __call__
    method = getattr(self, obj.__class__.__name__)
             ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
AttributeError: 'EntryPrinter' object has no attribute 'tuple'

bean-query --version
beanquery 0.3.0.dev0, beancount 2.3.6

[blais]

Works for me.

$ bean-query  --version
beanquery 0.1.dev1, beancount 3.2.0.dev0

[dnicolodi]

The reported error should be improved (this is still experimental functionality) but this is working as expected: the beancount output format can format only beancount objects, namely ledger entries: transactions, open and close statements, price statements, etc... There can be only one such object for row in the results set.

Your query is implicitly on the postings table. Each row of the postings table has an entry column that returns the transaction to which the posting belongs.

Your query works when written like this:

beanquery> .format beancount
beanquery> SELECT entry  WHERE account ~ "^Expenses:" 

Most likely, you don't want one transaction for each posting matching the WHRE clause (for transactions with more than one postings to an account matching ^Expenses:) thus you may want to add a DISTINCT clause:

beanquery> .format beancount
beanquery> SELECT DISTINCT entry  WHERE account ~ "^Expenses:" 

[tbm]

Thanks for your explanation. That was indeed not obvious.

[Maybe the following should be in a separate ticket as a "feature request", as opposed to this ticket to clarify the error message]

While I understand your explanation and rationale, I still believe that my query should "just work". Ok, maybe I should give some background to explain where I'm coming from. In ledger, you can do queries and reg will show the register output. Sometimes, the register output doesn't give me all the information I want, so then I simply change reg to print and it shows me the full entry that matches my query.

So while I understand that DISTINCT entry will solve the problem, that means I have to change my query compared to the "register" query I did initially. I was hoping I could just do .format beancount to get the ledger print equivalent while running exactly the same query as before.

Maybe the solution for me is to write a small script that changes the query accordingly, or maybe my use case makes some sense and can be implemented somehow.

[dnicolodi]

The output formatter is decoupled from the query engine and from the query parser. In principle, it would be possible to have the beanquery shell rewrite the query to do what you expect when the beancount output format is selected, but that would require special casing the structure of beancount tables and related queries even more into the beanquery shell, and I am trying to go into the opposite direction.

Your expectation makes sense coming from ledger but it does not when considering .format to change the representation of the query results. From this point of view it would be surprising if a query selecting date and meta("receipt") would return unrelated fields. At the very minimum, I would expect it to return only the receipt metadata field and not other metadata. However, this would require very complex query surgery to work.

[tbm]

Your expectation makes sense coming from ledger but it does not when considering .format to change the representation of the query results.

Yes, I think you're right. I need to play with bean-query more and get used to it.

Anyway, the error message should be improved, though, so let's leave this ticket open for that. Thank you for bean-query!