Inconsistent Decimal Mark Handling after CSV Import with Rules

[yse]

Environment:

• hledger version: 1.32.1-g39fc78a64-20231207, mac-x86_64

Steps to Reproduce:

1. Obtain the sample.csv and sample.csv.rules files from the examples directory on GitHub.
2. Import transactions from sample.csv using the command:

hledger import --rules-file=sample.csv.rules --file transaction.ledger sample.csv

3. Append a transaction with higher precision decimals to sample.csv:

echo '"2024/4/1","blah","-10000.00"' >> sample.csv

4. Verify the transaction in transaction.ledger. The decimal mark appears correctly:

2024-04-01 blah
    assets:bank:checking      $-10000.00
    expenses:unknown           $10000.00

5. Manually add a transaction with a comma as the thousand separator in transaction.ledger:

cat <<EOF >> transaction.ledger
2024-04-14 blahblah blah
    assets:bank:checking    \$2,000.00
    income:unknown          \$-2,000.00
EOF

6. Append another transaction to sample.csv:

echo '"2024/4/15","blah blah","-2000.00"' >> sample.csv

7. Re-import the updated CSV:

hledger import --rules-file=sample.csv.rules --file transaction.ledger sample.csv

8. Review the latest transaction in transaction.ledger and observe the incorrect and unexpected use of a comma as the thousand separator:

2024-04-15 blah blah
    assets:bank:checking      $-2,000.00
    expenses:unknown           $2,000.00

Expected Behavior:

The decimal mark handling should remain consistent across manual entries and CSV imports, respecting the formatting defined in the rules file or the default behavior of the system.

Actual Behavior:

Transactions imported from the CSV file after manually appending entries with a different decimal format in the ledger file show an unexpected change in the decimal mark format (introduction of a comma as the thousand separator).

[simonmichael]

Thanks for reporting!

Is the behaviour explained and justified by this:

import tries to follow the journal's existing commodity styles when it creates new entries.
During the first import, the journal contains no thousand commas, so the style for $ is no commas.
During the second import, because of the one thousand comma you added to the journal, the style with thousand commas is inferred for $, and used for the new entry.

Either avoiding adding unwanted thousand commas in the journal, or declaring the desired style explicitly in the journal with

commodity $1000.00

will prevent it.

[yse]

Thanks a lot for clarifying. I've just used only D directive. Commodity also specify style for reporting which is impressive.

Thank you for this remarkably beautiful and useful tool.