Do histograms support custom bin sizes (besides a single range)?

[mdouds]

For example, if I want a dashboard histogram chart for span durations, I can use the below dql to show bins of 100ms each:

fetch spans
| summarize count(), by:{range(duration, 100ms)}

Instead, can we create custom buckets, like 0-250ms, 250-500ms, 500ms-1s, 1s - 3s, etc.?

[t_pawlak · Accepted solution]

Yes — but not with range(). range(x, step) only makes equal-width bins. For custom buckets (0–250 ms, 250–500 ms, 500 ms–1 s, 1–3 s, …), build the buckets yourself via conditions and then group by that label.

Try this:

fetch spans
| fieldsAdd
  bucket =
    if(duration < 250ms, then: "0–250 ms", else:
      if(duration < 500ms, then: "250–500 ms", else:
        if(duration < 1s,   then: "500 ms–1 s", else:
          if(duration < 3s, then: "1–3 s", else: "≥3 s")
        )
      )
    ),
  bucket_idx =
    if(duration < 250ms, then: 1, else:
      if(duration < 500ms, then: 2, else:
        if(duration < 1s,   then: 3, else:
          if(duration < 3s, then: 4, else: 5)
        )
      )
    )
| summarize spans = count(), by: { bucket, bucket_idx }
| sort bucket_idx asc
| fieldsRemove bucket_idx

you should recive this:

[mdouds]

Awesome, thanks! I started exploring this same method, but instead of using a separate `bucket_idx` to handle the sorting, I only used millisecond-based buckets with a mix of substring and type conversion, like this:

fieldsAdd firstNum = toLong(substring(bucket,from:0,to:indexOf(bucket,"ms")))
| summarize by:{bucket, firstNum}, count()
| sort firstNum
| fieldsRemove firstNum

Your approach is great because I can use any unit in my bucket names.