historical: Beancount can handle commas in amounts.

And having it looks nicer, is more consistent with our historical
books, is less code for me, and is no more trouble for the user.

oxrlib/commands/historical.py

@@ -176,13 +176,13 @@ class BeancountFormatter(LedgerFormatter):
     COST_FMT = '{{{}}}'
 
     def __init__(self, cost_rates, price_rates=None,
-                 signed_currencies=(), base_fmt='###0.###',
+                 signed_currencies=(), base_fmt='#,##0.###',
                  rate_precision=5, denomination=None):
         super().__init__(
             cost_rates,
             price_rates,
             (),
-            base_fmt.replace(',', ''),
+            base_fmt,
             rate_precision,
             denomination,
         )

tests/test_historical.py

@@ -95,7 +95,6 @@ def check_fx_amount(config, lines, amount, cost, fx_code, fx_sign=None, price=No
         cost_re = '{{={}}}'.format(rate_fmt.format(cost))
         price_re = ' @ {}'.format(rate_fmt.format(price))
     else:
-        amount = amount.replace(',', '')
         cost_re = '{{{}}}'.format(rate_fmt.format(cost))
         if config.args.from_date is None:
             price_re = ''
@@ -115,7 +114,7 @@ def check_nonfx_amount(config, lines, amount, code=None, sign=None):
         else:
             expected = f'{amount} {code}\n'
     else:
-        expected = f'{amount.replace(",", "")} {code or "USD"}\n'
+        expected = f'{amount} {code or "USD"}\n'
     assert next(lines, "<EOF>") == expected
 
 def test_rate_list(single_responder, output, any_date):