Missing Values

[randomboolean]

New Functionalities

The proposed enhancements extend the FILL operation to support univariate operations, filling missing timestamps within individual GTS based on specified gaps or empty buckets.

This approach fills existing functionality gaps:

• the inability of bucketize fillers to utilize filler functions or to fill beyond empty buckets.
• the inability of existing filler functions and FILL framework to be used beyond cross filling use cases.

Function Signatures

• Existing Signature: Maintains the original functionality, aligning two GTS by filling to match their timestamps.

gtsa gtsb filler FILL

• New Signatures: Introduce univariate filling, with optional parameter for specific gaps (ticks) or empty buckets.

[ gtsList filler ] FILL

[ gtsList filler ticks ] FILL

gtsList {
  'filler' filler
  // optional:
  'ticks' ticks
  'invalid.value' NULL
} FILL

Note that the optional parameter invalid.value is the value returned by the evaluation function if the input timestamp is not in the function's support.

Additional Fillers

• filler.linear
• filler.spline
• filler.akima
• filler.newton
• filler.lowess
• filler.rlowess

MACROFILLER

When used in the context of the new univariate FILL mode, the macro encapsulated by MACROFILLER should be presented only with elements from the specified GTS, unlike in cross fill operations where it also receives elements from another GTS. The documentation will be revised to reflect this usage.

Macros as Filler Parameter

We have introduced the capability to use a macro directly as a filler parameter, (similar to a lambda filler function). In this setup, the macro must be designed to expect a GTS and a tick, and to return the filler value.

[pi-r-p]

To discuss: deduplication issues.
It can just fail, or deduplicate silently:

• with the full DEDUP logic ?
• with a SORT ?
  Failing is simpler, and dedups are easy to check after sort.