Closed danauerbach closed 6 years ago
Another way to describe this issue: dataselect is not able to trim/remove data from the middle of a record, leaving the earliest and latest samples.
To perform that operation requires copying the record and trimming it once for earliest samples and once for the latest samples. This is the simplest of the general case of multiple selections from a single record, which quite quickly becomes pathological. The work around is to do a single selection per run of dataselect, might have to be good enough for a while.
A test for this condition and printed warning are included in version 3.20 (https://github.com/iris-edu/dataselect/releases/tag/v3.20). A CAVEATS AND LIMITATIONS section was added to the documentation describing the condition and workaround.
When two selection criteria are within a single ms record, dataselect is not performing any selection.
For example, using the selection criteria:
II SUR 00 LN1 2017,187,21:35:00 2017,187,23:07:00 II SUR 00 LN1 2017,187,23:17:00 2017,188,00:00:00
when the end of the first selection (2017,187,23:07:00) and the start of the second (2017,187,23:17:00) exist within the a single record the complete record is passed to the output.
3 files are zipped and attached that include a selection file (select-ln1) with 3 selections, an input file (input-ln1.ms), and the resulting output file (output-ln1.ms).
dataselect-multi-select-example.zip
dataselect version 3.19 built and run on RHEL 6.
From email exchange with Chad (2017-07-12):