Data Explorer

Open a data file of any size and actually understand it. This is the tile for data profiling, the data quality score, the data dictionary, and SQL over a CSV — all four live here, in three tabs.

Data Explorer is read-only. It never changes your file, and it uploads nothing.

What it accepts. CSV, TSV, PSV and other delimited text; Excel workbooks (.xlsx, .xlsm, .xls, .xlsb); Parquet files and folders of Parquet parts; JSON and NDJSON.

How to use it

1. Open Data Explorer from the sidebar.
2. Drop a file onto Drop a data file here, or click Choose a file.
3. The file bar appears across the top and stays there. It shows what you are looking at — the file, the detected dialect, the sheet if it is a workbook — no matter which tab you are on.
4. Switch between the three tabs: Rows, Columns and SQL.

The file is opened once, and all three tabs share that one open file. Switching tabs is instant and never re-reads anything.

Data Explorer → Rows

The grid. Scroll through your data a page at a time, sort by a column, and filter. Only the rows on screen are ever read, which is why the size of the file does not decide whether you can look at it.

Paging through a file in its natural order is a fresh read each time you jump a long way; once you sort by a column, paging becomes cheap. Both are correct — the second is just faster on a distant jump.

Data Explorer → Columns

This is the profiling tab, and it is where a visitor arriving from a data-profiling or data-quality page should go.

For every column it reports the inferred type, how many values are missing, how many are distinct, the most common values, and — for numbers and dates — the minimum, maximum and average.

At the top sits the data quality score: a single 0–100 headline made of five weighted parts, with the weights shown, because an unexplained score is worse than no score.

Underneath, DataTray lists the specific issues it found, in plain language, each with a Show me these rows button that takes you to Data Explorer → Rows with a filter already applied. Finding a problem and seeing the rows that cause it are the same gesture.

The data dictionary. From this tab, Export writes a data dictionary — one row per column, with its name, type, null count, distinct count, examples and the role DataTray inferred (identifier, date, category, measure, free text) — as a spreadsheet you can hand to whoever asked what is in the file. Report writes the full data profile as a single HTML file. Both land in the Explorer subfolder and appear in the Job Queue.

Data Explorer → SQL

A plain SQL editor over the open file, for anyone who would rather write a query than click. This is the destination for anyone who came looking for SQL over a CSV.

• The tab is read-only by design. Only reading statements are accepted; anything that would write, attach a database, change a session setting or install an extension is refused before it reaches the engine, with a message saying so. That check is not a warning you can click past.
• Drag a file from Explorer into the editor to paste its full path.
• Save a view to give a query a name, then use that name in later queries.
• Results are capped at 1,000 rows on screen; Export writes the full result set to a file.
• Long queries can be cancelled while they run.

Output

Exports and reports are saved to the Explorer subfolder of your output folder.

If something goes wrong here

• The file opened as one column. The delimiter was detected wrongly. Change it in the file bar — delimiter, quote character, encoding, whether there is a header row — and the file reopens with your settings. The detected values are always shown, never applied silently.
• "DataTray ran out of memory reading this file." The run exceeded the memory budget rather than dragging your whole machine down. Close other jobs and try again, or work on a smaller selection.
• A workbook opened on the wrong sheet. Pick the sheet you want in the file bar. A multi-sheet workbook is never an error and never silently truncated to sheet one.
• Duplicate column headers are not reported. Known limitation in this release: if your CSV has two columns with the same name, the reader makes them unique (region, region_1) before the profiler sees them, so the duplicate column names issue never appears in this tab. Clean & Anonymize reads the raw header row and does report it.
• I cannot find a short alias for my file in the SQL tab. There isn't one — paste the full path (drag the file in), or save a named view and use that.