Simplifying a time-series segmentation Query

Question

This query segments a time-series and filters out dropout frames.

Can it be simplified - ie, made more readable?

segmentationQuery = 
 Query[SplitBy[#, NumberQ[#"POR_X"] &] &  ] /* 
  Query[Merge[Join] /* Key[True], GroupBy[NumberQ[#"POR_X"] &]]

(Note there's no SplitBy operator form in 10.0.0. hence using functional form.)

The Slot "POR_X" is used twice: in SplitBy and GroupBy - this is not pleasing.

Using only GroupBy in the top-most subquery yields only 2 groups of data (signal and blank) but all the numeric frames are merged (not the intended behavior).

This fragment, extracted from eye-tracking dataset, illustrates droputs of variable duration and number of rows:

data = {<|"Time" -> 0.033452, "POR_X" -> 613.76, 
  "POR_Y" -> 424.74|>, <|"Time" -> 0.066747, "POR_X" -> 611.84, 
  "POR_Y" -> 424.84|>, <|"Time" -> 0.100124, "POR_X" -> 610.53, 
  "POR_Y" -> 420.11|>, <|"Time" -> 0.133515, "POR_X" -> 605.95, 
  "POR_Y" -> 421.18|>, <|"Time" -> 2.01096, "POR_X" -> "", 
  "POR_Y" -> ""|>, <|"Time" -> 2.04425, "POR_X" -> "", 
  "POR_Y" -> ""|>, <|"Time" -> 2.07772, "POR_X" -> "", 
  "POR_Y" -> ""|>, <|"Time" -> 2.24451, "POR_X" -> 412.47, 
  "POR_Y" -> 805.92|>, <|"Time" -> 2.27792, "POR_X" -> 518.11, 
  "POR_Y" -> 652.38|>} //Dataset;

data[segmentationQuery][All, Dataset] // Normal

Answer 1

The following expression is slightly shorter:

data[
   SplitBy[#, NumberQ[#"POR_X"]&] &
][ Select[#[[1, "POR_X"]] =!= "" &], Dataset
] // Normal

Or, expressed as a Query:

query2 =
  Query[SplitBy[#, NumberQ[#"POR_X"]&] &] /*
  Query[Select[#[[1, "POR_X"]] =!= "" &], Dataset];

Answer 2

Small advice: use Missing[] instead of whitespace character. And dataset look better

PS: if you have deal with time series probably better use TemporalData or TimeSeries