Using flag-ms.py without a BITFLAG column (spoiler: it isn't possible [yet!])

[IanHeywood]

I want to do a very simple flagging operation on a MS, namely flagging a single timeslot. Here's what I'm trying:

flag-ms.py --flag --fill-legacy --timeslots=893  ../1612141271_sdp_l0_1024ch_scan10.ms

This produces this message:

flag-ms.py: MS does not contain a BITFLAG column. Please run the addbitflagcol utility on this MS.
Perhaps you forgot the -c/--create option?

Is there a way to use flag-ms.py for simple stuff like this without having to add a BITFLAG column?

Thanks.

[IanHeywood]

Note for people running this on computers they don't own: the addbitflagcol utility is available in the Stimela MeqTrees container.

I tried adding the BITFLAG column with flag-ms.py directly, which this issue suggests is possible:

flag-ms.py --create --init-bitflags 0 ../1612141271_sdp_l0_1024ch_scan10.ms
===> No actions were performed. Showing the result of your selection:
===>   MS size:                1869920 rows
===>   Data/time selection:    1869920 rows, 7659192320 visibilities (100% of MS rows)

but that didn't do anything.

[o-smirnov]

Try 16 or 32 instead of 0?

[IanHeywood]

That's worked (although I went for 0 because it's the default), but --flag is expecting a flagset:

flag-ms.py --flag FLAG --fill-legacy --timeslots=893  ../1612141271_sdp_l0_1024ch_scan10.ms
flag-ms.py: Flagset 'FLAG' not found

which worries me that I'm dangerously close to accidentally using bitflags again.

[o-smirnov]

which worries me that I'm dangerously close to accidentally using bitflags again.

Too late, if you're initializing a BITFLAG column, then you are already using bitflags quite purposefully. Now just add -c and your journey is complete. :) The regular FLAG column (aka legacy flags) should be populated by default anyway.

[IanHeywood]

You might have to hold my hand here.

flag-ms.py -c --flag FLAG --fill-legacy --timeslots=893  ../1612141271_sdp_l0_1024ch_MTP0013_scan10.ms
flag-ms.py: Flagset '--timeslots=893' not found

[o-smirnov]

--fill-legacy expects a flagmask and treats the next argument as such. But if you omit --fill-legacy, it will do the right thing by default.

[IanHeywood]

Bingo. Thanks.

flag-ms.py -c --flag FLAG --timeslots=893  ../1612141271_sdp_l0_1024ch_MTP0013_scan10.ms
  ===> select timeslots slice(893, 894, None) (reltime 1785.24~1785.24 s)
===> flagging with flagmask 0x0001
===> filling legacy flags with flagmask 0xFFFF
===> Flagging stats:
===>   MS size:                1869920 rows
===>   Data/time selection:       2080 rows,    8519680 visibilities (0.111% of MS rows)
===>     (over which legacy flags were filled using flagmask 0xFFFF)
===>     (which were flagged using flagmask 0x0001)

[o-smirnov]

Let's keep this open. You have a workaround, but it would be nice to do this directly on legacy flags without having to insert bitflags.