CSV profiler
On-deviceThis page always downloads the big-file engine, even for a small file — the engine is what does the counting. It downloads once and then stays on your device.
The profiler opens a CSV with a database engine that reads it from disk, so a file too large for a spreadsheet still gets counted. Every figure on this page — rows, empty cells, distinct values, smallest and largest — is counted over every row in the file. The column type is the only thing the page infers, and it says so beside it. Nothing about your file is uploaded.
About this tool
What you provide
- One CSV or TSV file.
What you get
- The row count, and for every column: how many values are filled, how many are empty, how many distinct values, and the smallest and largest value.
- The inferred type of every column, labelled as an inference and naming the sample it was inferred from.
Limits
- Every count is exact and computed over the whole file — nothing is approximated or sampled.
- The column type is inferred from only the first 20,480 rows of the file; if a column changes shape after that, you can re-read the whole file as text instead.
- Reading the file needs a one-time database-engine download of about 7 MB, downloaded even for a small file because the engine is what does the counting.
- Everything runs on your device and nothing about your file is uploaded.
How to use it
-
Choose a CSV file
Tap “Choose a CSV file” or drop a .csv or .tsv file onto the dashed area.
-
Download the engine
Tap “Download engine (~7 MB · one time)” — it downloads once and stays on your device for every file after that.
-
Profile the file
Tap “Profile this file” and wait while every row in the file is counted.
-
Read the results
See the row count and, for every column, the filled/empty/distinct counts and the smallest and largest value, plus the inferred type and the sample it came from.
Frequently asked questions
Why does this page ask me to download an engine even for a small file?
The profiler always uses the same database engine to count your file, whatever its size — the download happens once, then stays on your device for every file after that.
Are the row and distinct-value counts exact, or approximate?
Exact. Every figure on this page — rows, filled cells, empty cells, distinct values, smallest and largest — is counted over the whole file, never sampled or estimated.
Is the column type shown here always correct?
No — it is inferred from only the first 20,480 rows of the file. If a column changes shape further down (numbers for a long stretch, then text), the type can be wrong; you can re-read the file with every column as text instead.
Does profiling my file upload it anywhere?
No. The file is read on your device by the downloaded engine, and nothing about it is sent to a server.