Time-Series Resample
Bucket a timestamped CSV into 15-minute, hourly, daily, weekly, monthly, quarterly or yearly intervals and aggregate every value column — mean, sum, min/max, count, median, std/var or OHLC candles. Gap filling, bucket labelling and grid offsets included. Runs entirely in your browser, no upload.
About this tool
Resampling is the time-series version of "group by": every row is dropped into a fixed time bucket — 15 minutes, an hour, a day, a week, a month, a quarter, a year — and the values inside each bucket are collapsed into one number. It is how minute-by-minute sensor readings become an hourly chart, how daily orders become weekly totals, and how raw trades become OHLC candles.
Paste the CSV, pick an interval, pick an aggregate. Everything runs locally in your browser via WebAssembly — the series is never uploaded.
Worked example
Input (half-hourly readings) with interval 1h and aggregate mean:
time,temp
2024-05-01T10:00:00Z,10
2024-05-01T10:30:00Z,20
2024-05-01T11:15:00Z,30
Output — the 10:00 bucket averages 10 and 20, the 11:00 bucket holds only 30:
time,temp
2024-05-01T10:00:00Z,15
2024-05-01T11:00:00Z,30
Switch aggregate to ohlc and each value column expands into four:
temp_open,temp_high,temp_low,temp_close.
What it accepts
- Delimiters: comma, tab, semicolon or pipe — auto-detected from the first line and reused in the CSV output. Quoted cells with embedded delimiters are handled.
- Headers: optional. A first row that is entirely timestamps/numbers is treated as
data, and the output columns are named
time,column2,column3… - Timestamps: ISO-8601 / RFC-3339 (
2024-05-01T13:20:00Z, with or without an offset — offsets are converted to UTC),2024-05-01 13:20, plain2024-05-01, and bare epoch numbers (values of 1e11 or more are read as milliseconds, smaller ones as seconds). - Row order: any. Rows are sorted before bucketing, and buckets come back chronologically.
- Value columns: by default every column beside the timestamp whose cells are all
numeric or blank. Name specific ones with
value_columns(header names or 1-based numbers). Blank cells are ignored rather than counted as zero.
Limits
Input is capped at 2,000,000 bytes and 200,000 data rows, and the result is capped
at 100,000 buckets — a very fine interval over a long span hits that cap and asks for a
coarser one. Everything is computed in UTC; use offset to move the day boundary.
Upsampling
An interval finer than the data works too. With fill set to anything but skip, the
buckets between two rows are created and then filled — previous holds the last value,
linear interpolates. Daily 10 and 30 two days apart become 10, 20, 30 hourly-style
at 1d/linear.
FAQ
What is the difference between `label` and `closed`?
closed decides which rows land in a bucket; label only decides what timestamp is
printed. With closed = left (the default) a bucket covers [start, end), so a row exactly
on an edge opens the new bucket. With closed = right it covers (start, end], so an
exact-edge row closes the previous one. label = end prints the closing edge instead of the
opening one, which is what many reporting tools expect for "week ending" columns.
How do I get days that start at midnight in my own timezone?
Set offset to your UTC difference. offset = -5h with interval = 1d makes each day run
from 05:00 UTC to 05:00 UTC, which is midnight-to-midnight at UTC-5. The tool has no
timezone database and does not apply daylight-saving transitions — the shift is a fixed
duration, so pick the offset that matches the period you are summarising.
What does `origin` change?
It moves the whole bucket grid. epoch (the default) anchors edges to the Unix epoch, so
hourly buckets start on the hour and weekly buckets start on a Monday. start anchors them
to the first row's exact timestamp, so a series beginning at 10:20 gets buckets at 10:20,
11:20, … start_day anchors to UTC midnight of the first row's day. Month, quarter and year
buckets always start on the 1st, so origin does not affect them.
What happens to intervals with no data?
By default (fill = skip) they simply do not appear — the output only contains buckets that
had rows. empty emits the bucket with blank values, zero writes 0, previous carries the
last known value forward, and linear interpolates between the values on either side of the
gap. linear leaves a leading or trailing gap blank, because there is nothing on one side to
interpolate from.
Which aggregate should I use for counting events?
count — it reports how many rows in the bucket had a number in that column, so blank cells
are not counted. If your rows are events with no numeric payload, add a column of 1s and
use sum, or point value_columns at any always-populated numeric column and use count.
Are `std` and `var` sample or population statistics?
Sample — they use the n-1 denominator, the same default spreadsheets and pandas use. A
bucket holding a single value has no sample spread, so its cell comes back blank rather than
0.
Can I aggregate several columns at once?
Yes. Leave value_columns blank and every numeric column is aggregated with the chosen
function, each keeping its own header name. With ohlc, each of those columns expands into
four (<name>_open, <name>_high, <name>_low, <name>_close). Applying different
functions to different columns in one pass is not supported — run the tool once per function.
Is my data uploaded anywhere?
No. The resampler is compiled to WebAssembly and runs inside your browser tab. The CSV you paste never leaves the page, and the same engine is available offline through the command line.
Developer & Automation Access
Run it from the terminal
Same engine as this page, headless — via the gizza CLI:
gizza tool time-series-resample "time,temp
2024-05-01T10:00:00Z,10
2024-05-01T10:30:00Z,20
2024-05-01T11:15:00Z,30"New to the CLI? Get gizza →
Open it by URL
Pre-fill and auto-run this tool with query parameters — the names match the API/CLI:
https://gizza.ai/tools/time-series-resample/?data=time%2Ctemp%0A2024-05-01T10%3A00%3A00Z%2C10%0A2024-05-01T10%3A30%3A00Z%2C20%0A2024-05-01T11%3A15%3A00Z%2C30&interval=1h&aggregate=mean&time_column=blank%20%3D%20first%20column&value_columns=blank%20%3D%20every%20numeric%20column&label=start&closed=left&fill=skip&origin=epoch&offset=blank%20%3D%20no%20shift&time_format=iso&output=csvMachine-readable descriptor: tool.json — title + parameters JSON Schema for agents.