Bank CSV support request: build a synthetic failing example

The desktop apps export date, description, and amount as CSV. Open CSV in Excel to save an XLSX workbook. Statement support varies: verify the result against your own PDF.

Working sequence

Create a fully invented fixture rather than editing only the account name in a real statement.
Retain the failing delimiter, quote and embedded-newline structure while replacing every real value.
Run the same parser and record its settings, expected records/fields and actual result.
Compare a negative control; if simplifying the descriptions removes the failure, restore the necessary fictional structure before reporting it.

Preserve the mechanism, replace the records

Record the program and version, selected encoding, delimiter, quote character, expected columns and observed error. Follow Mozilla’s bug-writing guidance: give reproducible steps and distinguish expected from actual behavior. A screenshot of a broken table alone does not preserve the input bytes.

For a delimiter problem, keep a delimiter inside an invented quoted description. For a record-boundary problem, retain an embedded newline. If the original issue involves decimal commas, preserve that convention with invented amounts. Remove real account identities, references, paths and free-text narratives entirely; recreating a mechanism is different from claiming anonymization.

A smaller sample can accidentally stop failing

Our invented semicolon fixture has three fields per record:

date;description;amount
2026-02-03;"SYNTHETIC shop; note";"-12,30"
2026-02-04;"SYNTHETIC line one
line two";"4,00"

We actually read these bytes using Python 3.13.7 and csv.reader, with delimiter=';' and newline=''. The result has three records including the header, each with three fields. A physical-line split instead makes four lines and breaks the quoted description. Choosing a comma delimiter also produces the wrong field structure.

Replacing both descriptions with SYNTHETIC removes the embedded delimiter and newline. A naive splitter can then appear correct: that simplified sample no longer demonstrates the original mechanism. Keep the negative control alongside the successful parse. These are Python parser results, not an Excel or accounting-import test.

Send a bounded question

Use a short request such as: “Python 3.13.7, semicolon delimiter, three expected fields. The attached invented input has a quoted newline. Splitting physical lines gives four lines; csv.reader gives three records. Which parser setting should this importer use?” Include only the invented artifact, relevant settings and error text checked for private paths or identifiers.

Before sharing, reopen the exact attachment, inspect every row and compare its bytes with the reviewed example. Check the recipient and approved channel separately. No diagnostic success proves that an AI service’s retention terms or a support upload is authorized. For fixing the real import after diagnosis, use the existing delimiter/encoding guide rather than copying the client source into a public prompt.

Observed reference controls

Executed teaching control	Observed records/fields	Diagnostic meaning
Semicolon csv.reader with newline handling	3 logical records,3 fields each	Expected structural reading
Wrong comma delimiter	Not 3 fields in every record	Delimiter symptom remains
Physical-line splitting	4 lines, ragged fields	Embedded newline breaks a naïve line-based reader
Over-simplified fictional descriptions	3 lines,3 fields each under line splitting	Simplification erases the original symptom

Synthetic review case. The values are teaching data, not a vendor export or an import file.

Optional worksheet starter.

For a bank source available only as PDF, download LocalExtract. It processes one PDF locally into date,description,amount CSV; check the extracted rows against the issued statement before using them.

Documentation for the added task checked October 16, 2026.