Add UNCONFIRMED to externally predicted postings of imported entries

[dumbPy]

If a posting was predicted by smart_importer or was entered by hand, it won't be predicted again but should be editable like the FIXME or predicted entry by beancount-import.

This was suggested and discussed in #60 and will start working once #60 is completed and merged.

[coveralls]

Pull Request Test Coverage Report for Build 222

• 29 of 29 (100.0%) changed or added relevant lines in 2 files are covered.
• No unchanged relevant lines lost coverage.
• Overall coverage increased (+0.1%) to 68.986%

---

Totals
Change from base Build 216:	0.1%
Covered Lines:	5160
Relevant Lines:	7245

---

💛 - Coveralls

[jbms]

Please change variable name to UNCONFIRMED

[Zburatorul]

@dumbPy, glad you went ahead with this. How do I set up the minimal example to test that #62 works?

[dumbPy]

Edit:

You may use the test_dummy_predictor test added in 50c6ba9 to test the input and working of this PR.

Original Comment (outdated):

@Zburatorul use below code as generic_importer_source_test.py. It will add dummy postings to all the imported transactions just like the smart_importer. the basic_test fails with it, but the other two tests pass.

You can also use the DummyPostingPredictor from below with any other importer and statements you have, and it will add the balanced Dummy posting. #82 adds {"unconfirmed_account":""} to meta of this Dummy Posting. #60 should make this posting editable from frontend.

from decimal import Decimal
import os

import pytest

from .source_test import check_source_example
from beancount.ingest.importers.csv import Importer as CSVImporter, Col
from beancount.core.data import Transaction, Posting, Amount

testdata_dir = os.path.realpath(
    os.path.join(
        os.path.dirname(__file__), '..', '..', 'testdata', 'source', 'generic_importer'))

testdata_csv = os.path.join(testdata_dir, "csv")

examples = [
    'test_basic',
    'test_invalid',
    'test_training_examples'
]

importer = CSVImporter({Col.DATE: 'Date',
                        Col.NARRATION1: 'Description',
                        Col.AMOUNT: 'Amount',
                        },
                       'Assets:Bank',
                       'USD',
                       '"Date","Description","Amount"',
                       )

class DummyPostingPredictor:
    DUMMY_ACCOUNT = "Assets:Dummy"

    def __init__(self, importer):
        self.importer = importer
        # move extract to _extract
        self.importer._extract = self.importer.extract
        self.importer.extract = self.extract_wrapper
    
    def extract_wrapper(self, f, existing_entries):
        entries = self.importer._extract(f, existing_entries)
        for entry in entries:
            if isinstance(entry, Transaction):
                p = entry.postings[0]
                entry.postings.append(
                    Posting(
                        account=self.DUMMY_ACCOUNT,
                        units=Amount(currency=p.units.currency, number=-1*p.units.number),
                        cost=None,
                        price=None,
                        flag=None,
                        meta={},
                    )
                )
        return entries

# wrap the importer with the dummy posting predictor.
# Will predict 2nd dummy postings for the importer entries
DummyPostingPredictor(importer)

@pytest.mark.parametrize('name', examples)
def test_source(name: str):
    check_source_example(
        example_dir=os.path.join(testdata_dir, name),
        source_spec={
            'module': 'beancount_import.source.generic_importer_source',
            'directory': testdata_csv,
            'account': 'Assets:Bank',
            'importer': importer,
        },
        replacements=[(testdata_dir, '<testdata>')])

[jbms]

Perhaps use True rather than "" since this seems to be a boolean field?

[jbms]

Can you please add a unit test?

[dumbPy]

@jbms @Zburatorul Hey guys, this feels ready now.
Sorry I dropped out of the loop 4 years ago :p
Cherry-picked from #60
Now, if you wrap the importers in smart_importer's predictors,

for importer in CONFIG:
    apply_hooks(importer, [PredictPostings(), PredictPayees()])

1. The imported transactions are added posting by PredictPostings()
2. generic_importer_source.py adds "unconfirmed_account":True metadata to the posting.
3. reconcile.py checks for this "unconfirmed_account":True, pops the posting, and adds FIXME posting in it's place and uses the corresponding account as dummy prediction.
4. This dummy prediction is now passed forward as if it was predicted by the beancount-import's predictor allowing editing in the frontend.

[dumbPy]

@Zburatorul I have been using this too for a few months and works perfectly fine as expected.