[daip] VBGLU and sort order etc

[Eric Greisen]

Michael Bietenholz wrote:
> I just noticed that if you run VBGLU (31DEC09; yesterday) on two data
> sets, both of which are in strict 'TB' order (not 'T?' labelled as
> 'TB'), then the VBGLU output is not necessarily in 'TB' order.
> 
> I think this happens when both data sets are missing some visibilities
> 
> Here's PRTUV output that shows what happens:
> 
>   Vis #     IAT      Ant  Su Fq U(klam) V(klam) W(klam)  etc
> 
> Part 1
>    15292  0/16:19:27  2- 8  1  1   22127  -18513   26334
>                      (2- 9 is missing from part 1 at this time-stamp)
>    15293  0/16:19:27  2-10  1  1   28088    4454   15699
> 
> Part 2
>    15165  0/16:19:27  2- 8  1  1   22252  -18618   26482
>    15166  0/16:19:27  2- 9  1  1   24996  -15241   24813
>    15167  0/16:19:27  2-10  1  1   28247    4479   15787
> 
> Both Part 1 and Part 2 are in TB order, but after VBGLU we have this
> 
> VBGLU:
>    15292  0/16:19:27  2- 8  1  1   22127  -18513   26334
>    15293  0/16:19:27  2-10  1  1   28088    4454   15699
> ...
>    15330  0/16:19:27  2- 9  1  1   24855  -15156   24674
>     with the last vis. being now out of TB order.
> 
> So probably VBGLU matches up PT1 vis with their corresponding PT2
> bits, writes them out, and then processes any missed bits in PT2.
> 
> This isn't really a problem, only the sort order on output is wrongly
> set to 'TB'.  So it might be good if VBGLU set the sort-order on
> output to 'T' (or, more elaborately, checked for this condition,
> and set sort-order to 'T' only if order has got changed).
> 
> Also, perhaps VBGLU should write its AIPS version into the history?

Good point about the sort - I have added code to test the time and 
baseline sort order and control what is says on output.  The aips 
version (31DECxx) should appear in the HI after the copy of the old HI 
file from UV data set #1 since it calls HISCOP like many tasks.

Eric Greisen