Data Converter

One tile, three shapes of the same job: convert each, combine many into one, or split one into many.

What it reads. CSV, TSV, PSV and .txt; fixed-width text (when you describe the layout); Excel workbooks (.xlsx, .xlsm, .xls, .xlsb); Parquet; JSON, NDJSON and JSONL; XML.

What it writes. CSV, TSV, PSV, Excel (.xlsx), Parquet, JSON, NDJSON, XML, SQL INSERT statements, and an HTML table.

Fixed-width is a source format only — DataTray can read a fixed-width file but does not write one, and the panel refuses that target up front rather than letting you queue a run that cannot finish.

How to use it

1. Open Data Converter.
2. Drop your files in, or click to browse. Add as many as you like.
3. Choose the cardinality: convert each · combine · split.
4. Choose the target format and its options.
5. Press the run button.

The dialect is shown, not guessed silently

The single most common complaint about any CSV tool is a file that opened as one column. DataTray reads a sample of your file, shows you what it concluded — delimiter, quote character, escape character, encoding, header row, rows to skip — and lets you change any part of it before the run. The guess is visible; that is the whole point.

Multi-sheet workbooks are not an error

Drop a forty-sheet workbook and DataTray lists all forty sheets and lets you pick. Select all of them and you get forty outputs, not sheet one silently.

Workbooks are read through a small conversion step first: DataTray writes the sheet you picked out to a temporary CSV in your temp folder, then reads that. It is why a very large workbook takes a moment to appear, and the temporary files are cleaned up when you quit.

Combine

Combining several files into one reconciles their columns first, and shows you the reconciliation before it runs: which columns appear in which files and what will be empty where. Two options matter:

• Union by name — line columns up by their names rather than their positions, so a file with an extra column in the middle still lands correctly.
• Add a source column — record which file each row came from.

Split

Split one file into many, either by a fixed number of rows per part, or by the distinct values of a column (all the West rows in one file, all the East rows in another).

Two honest notes about splitting by column:

• Splitting to CSV or Parquet is a single pass and is the fastest route.
• Splitting to JSON or NDJSON is slower, because the engine cannot partition those formats in one pass and DataTray has to write one file per distinct value in turn. It works; it just takes longer on a column with many distinct values.

CSV output options

Every CSV DataTray writes is, by default, UTF-8 with a byte-order mark and Windows (CRLF) line endings, which is what Excel on Windows needs in order to read accented characters correctly. If your Windows list separator is a semicolon, DataTray honours it. All of these are adjustable in the panel.

One exception, worth knowing: when you split by column to CSV, the individual part files inside the generated folders are written by the engine directly and do not carry the byte-order mark or the Windows line endings. If Excel shows mangled accents in one of those parts, that is why — run the part back through a plain convert to normalise it.

Parquet output options

Compression is selectable: zstd (the default), snappy, gzip or none.

Output

Saved to the Converted subfolder of your output folder, unless you choose a different destination in the panel.

If something goes wrong here

• My file opened as one column. Change the delimiter in the dialect strip and re-run.
• A target is greyed out. The pair is not a conversion — a source never offers its own format as a target, .xls and .xlsb do not offer .xlsx (converting one workbook format to another is a different job), and fixed-width is not offered as a target at all.
• The whole drop was rejected. Nothing in the selection carried an extension DataTray reads.
• Rows are missing from the output, and there is a rejects file. Working as designed: rows the reader could not parse are collected in … - rejects.csv with the line, the column and the reason, instead of quietly becoming blanks.
• A fixed-width file came out wrong. Fixed-width is the one format where bad values can be dropped without appearing in a rejects file. Double-check the field widths you entered.