Skip to main content
Data type Time64 represents a time-of-day with fractional seconds. It has no calendar date components (day, month, year). The precision parameter defines the number of fractional digits and therefore the tick size. Tick size (precision): 10-precision seconds. Valid range: 0..9. Common choices are 3 (milliseconds), 6 (microseconds), and 9 (nanoseconds). Syntax: 
Time64(precision)
Internally, Time64 stores a signed 64-bit decimal (Decimal64) number of fractional seconds. The tick resolution is determined by the precision parameter. Time zones are not supported: specifying a time zone with Time64 will throw an error. Unlike DateTime64, Time64 does not store a date component. See also Time. Text representation range: [-999:59:59.000, 999:59:59.999] for precision = 3. In general, the minimum is -999:59:59 and the maximum is 999:59:59 with up to precision fractional digits (for precision = 9, the minimum is -999:59:59.999999999).

​
Implementation details

Representation. Signed Decimal64 value counting fractional second with precision fractional digits. Normalization. When parsing strings to Time64, the time components are normalized and not validated. For example, 25:70:70 is interpreted as 26:11:10. Negative values. Leading minus signs are supported and preserved. Negative values typically arise from arithmetic operations on Time64 values. For Time64, negative inputs are preserved for both text (e.g., '-01:02:03.123') and numeric inputs (e.g., -3723.123). Saturation. The time-of-day component is capped to the range [-999:59:59.xxx, 999:59:59.xxx] when converting to components or serialising to text. The stored numeric value may exceed this range; however, any component extraction (hours, minutes, seconds) and textual representation use the saturated value. Time zones. Time64 does not support time zones. Specifying a time zone when creating a Time64 type or value throws an error. Likewise, attempts to apply or change the time zone on Time64 columns is not supported and results in an error.

​
Examples

  1. Creating a table with a Time64-type column and inserting data into it:
CREATE TABLE tab64
(
    `event_id` UInt8,
    `time` Time64(3)
)
ENGINE = TinyLog;

-- Parse Time64
-- - from string,
-- - from a number of seconds since 00:00:00 (fractional part according to precision).
INSERT INTO tab64 VALUES (1, '14:30:25'), (2, 52225.123), (3, '14:30:25');

SELECT * FROM tab64 ORDER BY event_id;

   ┌─event_id─┬────────time─┐
1. │        1 │ 14:30:25.000 │
2. │        2 │ 14:30:25.123 │
3. │        3 │ 14:30:25.000 │
   └──────────┴──────────────┘
  1. Filtering on Time64 values
SELECT * FROM tab64 WHERE time = toTime64('14:30:25', 3);

   ┌─event_id─┬────────time─┐
1. │        1 │ 14:30:25.000 │
2. │        3 │ 14:30:25.000 │
   └──────────┴──────────────┘

SELECT * FROM tab64 WHERE time = toTime64(52225.123, 3);

   ┌─event_id─┬────────time─┐
1. │        2 │ 14:30:25.123 │
   └──────────┴──────────────┘
Note: toTime64 parses numeric literals as seconds with a fractional part according to the specified precision, so provide the intended fractional digits explicitly.
  1. Inspecting the resulting type:
SELECT CAST('14:30:25.250' AS Time64(3)) AS column, toTypeName(column) AS type;

   ┌────────column─┬─type──────┐
1. │ 14:30:25.250 │ Time64(3) │
   └───────────────┴───────────┘

​
Addition with Date

A Time64 value can be added to a Date or Date32 value to produce a DateTime64 with the same scale as the Time64: 
SET use_legacy_to_time = 0;
SELECT toDate('2024-07-15') + toTime64('14:30:25.123456', 6) AS dt, toTypeName(dt);

   ┌─────────────────────────dt─┬─toTypeName(dt)─┐
1. │ 2024-07-15 14:30:25.123456 │ DateTime64(6)  │
   └────────────────────────────┴────────────────┘
See Date and Time Addition for details on all supported combinations and result types. See Also
Last modified on June 8, 2026