search

desihub / fastspecfit

Fast spectral synthesis and emission-line fitting of DESI spectra.
https://fastspecfit.readthedocs.org
BSD 3-Clause "New" or "Revised" License
12 stars 2 forks source link

photsys can be empty for some secondary targets #32

Closed moustakas closed 2 years ago

moustakas commented 3 years ago

Summarizing an off-list conversation with Raga, @geordie666, and @stephjuneau:

To get PHOTSYS from TARGETID (thanks to @geordie666):

from desitarget.io import releasedict
from desitarget.targets import decode_targetid  
_, _, releases, _, _, _ = decode_targetid(targets["TARGETID"])  
photsys = [releasedict[release] if release >= 9000 else None for release in releases]

This should return "N"/"S" as expected, or None if the target has no Legacy Surveys provenance (by which I mean that no valid TARGETID was propagated from dr9 for the target).

But then Raga found:

Adam's code works if the release is not equal to 11 - so, I am assuming it works for 
the targets that have DR9 photometry. 

Here are two targets for testing - 
1. TARGETID: 39633489506601735, TILEID: 80696, RELEASE: 9011, PHOTSYS: 'N'
2. TARGETID: 39632956502838508, TILEID: 80895, RELEASE: 9011, PHOTSYS: 'N'

There is another thing that I have found in the fibermaps. In some cases, the flux_* values 
are 0.0 - even when they have valid DR9 phtometry. And this is not always the case for just 
secondary targets. I have seen ~10 BGS targets with FLUX_G = 0.0 - I still need to investigate 
this further.  From the above two targets, the first target have FLUX_G, R, Z = 0 from the FIBERMAP - 
but, it has valid values in DR9 photometry.
Ragadeepika-Pucha commented 3 years ago

I am sorry for the delay. Tagging myself on this issue.

moustakas commented 2 years ago

I've confirmed that this is an issue and I think there's a straightforward fix.

The problem is that the following catalogs

/global/cfs/cdirs/desi/target/catalogs/dr9/0.50.0/targets/sv1/secondary/dark/sv1targets-dark-secondary.fits
/global/cfs/cdirs/desi/target/catalogs/dr9/0.50.0/targets/sv1/secondary/bright/sv1targets-bright-secondary.fits

only have a small fraction of the columns normally found in a targets catalog. Consequently, when fiberassign is/was run, the columns which are usually populated like FLUX_[G,R,Z], PHOTSYS, etc., are empty, e.g.,

from astropy.table import Table
info
<Table length=1>
     TARGETID        TARGET_RA        TARGET_DEC
      int64           float64          float64
----------------- ---------------- ----------------
39633489506601735 149.490652378136 70.0725222414694

fm = Table.read('/global/cfs/cdirs/desi/target/fiberassign/tiles/trunk/080/fiberassign-080696.fits.gz', 'FIBERASSIGN')
targ = Table.read('/global/cfs/cdirs/desi/target/fiberassign/tiles/trunk/080/fiberassign-080696.fits.gz', 'TARGETS')

fm = fm[fm['TARGETID'] == info['TARGETID']]
targ = targ[targ['TARGETID'] == info['TARGETID']]
fm['PHOTSYS', 'FLUX_G', 'FLUX_R', 'FLUX_Z']
<Table length=1>
PHOTSYS  FLUX_G  FLUX_R  FLUX_Z
 bytes1 float32 float32 float32
------- ------- ------- -------
            0.0     0.0     0.0

This particular target, however, has perfectly valid photometry from DR9:

import astropy.units as u
from astropy.coordinates import SkyCoord
tt = Table(fitsio.read('/global/cfs/cdirs/cosmo/data/legacysurvey/dr9/north/sweep/9.0/sweep-140p070-150p075.fits'))
c1 = SkyCoord(ra=tt['RA']*u.deg, dec=tt['DEC']*u.deg)
c2 = SkyCoord(ra=info['TARGET_RA']*u.deg, dec=info['TARGET_DEC']*u.deg)
indx, _, _ = c2.match_to_catalog_sky(c1)
tt[indx]['FLUX_G', 'FLUX_R', 'FLUX_Z', 'RELEASE']
<Table length=1>
  FLUX_G    FLUX_R   FLUX_Z  RELEASE
 float32   float32  float32   int16
--------- --------- -------- -------
1.6022577 4.1805983 9.720711    9011

It's this photometry, of course, which I need as input to fastspecfit to do my SED modeling.

My proposed solution is to create two new row-matched catalog to these two files

/global/cfs/cdirs/desi/target/catalogs/dr9/0.50.0/targets/sv1/secondary/dark/sv1targets-dark-secondary.fits
/global/cfs/cdirs/desi/target/catalogs/dr9/0.50.0/targets/sv1/secondary/bright/sv1targets-bright-secondary.fits

which contain the original DR9 photometry (plus a couple other columns normally generated by the target-selection code like PHOTSYS), and then I'll reach those catalogs for the handful of tiles which contain these secondary targets.

@geordie666 is this a reasonable solution?

If so, @stephjuneau or @Ragadeepika-Pucha would you be willing to generate these catalogs?

geordie666 commented 2 years ago

@moustakas: What you propose is certainly a reasonable solution for the small subset of cases where secondary targets (i) exist in DR9 and;(ii) were observed as part of sv1. For wider, general context:

moustakas commented 2 years ago

Thanks @geordie666, I'm not planning to support targets without DR9 photometry.

For the record here's my current procedure (please comment!):

  1. Given a single coadd-*.fits file and corresponding redrock-*.fits file, read the REDSHIFTS, FIBERMAP and EXP_FIBERMAP extensions.
  2. Apply quality cuts, currently: z>0 & zwarn <= 4 & objtype='TGT' & coadd_fiberstatus==0 & (photsys == 'N' | photsys == 'S').
  3. Read the header of the fiberassign file(s) for each unique TILEID in the EXP_FIBERMAP table. Use these headers to infer the target catalog directory(ies) used during fiber-assignment via the TARG, TARG2, TARG3, TARG4, SCND (if present) header cards.
  4. Read the following columns from the original target catalogs: ['TARGETID', 'RA', 'DEC', 'FIBERFLUX_G', 'FIBERFLUX_R', 'FIBERFLUX_Z', 'FIBERTOTFLUX_G', 'FIBERTOTFLUX_R', 'FIBERTOTFLUX_Z', 'FLUX_G', 'FLUX_R', 'FLUX_Z', 'FLUX_W1', 'FLUX_W2', 'FLUX_IVAR_G', 'FLUX_IVAR_R', 'FLUX_IVAR_Z', 'FLUX_IVAR_W1', 'FLUX_IVAR_W2']

The trouble with this sequence is that the SV1 targets with PHOTSYS blank (including the low-mass AGN that @Ragadeepika-Pucha and @stephjuneau are interested in) get dropped in step [2] because of the PHOTSYS cut.

So what I'd really like to do is to remove the PHOTSYS cut from step [2], read PHOTSYS in step [4] (after we've generated new row-matched photometric catalogs for the SV1 targets), and add an additional selection cut in or after step [4] to remove objects that don't have DR9 photometry (unless there's a way for me to glean that information in step [2]???).

geordie666 commented 2 years ago

I think that sequence is correct @moustakas. The only way I can think of to (potentially) simplify step (2) is to make an explicit check on whether a target is from sv1. If the target has photsys blank, but is from sv1 you would then retain the object for special treatment. It's really only sv1 that's the problem, here, I think.

There might be header information corresponding to sv1 in some of the files you're reading. But, certainly, if the target columns look like, e.g., SV1_DESI_TARGET in the coadd file, then its an sv1 file. To check whether targets are from sv1, you can use desitarget.targets.main_cmx_or_sv, e.g.:

import fitsio
from desitarget.targets import main_cmx_or_sv 
objs = fitsio.read("/global/cfs/cdirs/desi/spectro/redux/daily/tiles/cumulative/80865/20210510/coadd-3-80865-thru20210510.fits", "FIBERMAP")
cols, Mx, survey = main_cmx_or_sv(objs)
survey
Out[]: 'sv1'
Ragadeepika-Pucha commented 2 years ago

@moustakas - I can create the catalogs that you need for the SV1 secondary targets. Can you confirm what columns you need from DR9 photometry? Are these the ones from step [4]? I can add blank PHOTSYS and '-99' for the rest of the columns for the targets without available DR9 photometry. Or would you suggest something else?

moustakas commented 2 years ago

Terrific, @Ragadeepika-Pucha, thanks. If you could please propagate every column that's in the DR9 sweep files (plus the TARGETID that's in the targets catalog now and PHOTSYS), that way we won't have to do this again.

To determine which SV1 secondary targets should have DR9 photometry (rather than having to match on RA,Dec), I just checked that the trick @geordie666 gave us above works beautifully:

import numpy as np
from astropy.table import Table
from desitarget.targets import decode_targetid
tt = Table.read('/global/cfs/cdirs/desi/target/catalogs/dr9/0.50.0/targets/sv1/secondary/dark/sv1targets-dark-secondary.fits')
objid, brickid, releases, _, _, _ = decode_targetid(tt['TARGETID'])
print(np.count_nonzero(releases > 9000), len(tt))
  18882222 22354100

(And we can use objid and brickid to find the right row of the sweep catalog.)

Oh and let's use zero for empty column values (not -99). That's my preference at least. Thanks!

moustakas commented 2 years ago

@Ragadeepika-Pucha has prepared row-matched catalogs---thank you!

@geordie666 could you help review?

ioannis@cori09:~% ls -l /global/u1/r/raga19/data/
total 23530324
-rw-rw-r--+ 1 raga19 desi  2007959040 Aug 11 21:59 sv1targets-bright-secondary-dr9photometry.fits
-rw-rw-r--+ 1 raga19 desi 21236420160 Aug 13 10:51 sv1targets-dark-secondary-dr9photometry.fits

cd /global/u1/r/raga19/data/
fitsinfo sv1targets-dark-secondary-dr9photometry.fits
Filename: sv1targets-dark-secondary-dr9photometry.fits
No.    Name      Ver    Type      Cards   Dimensions   Format
  0  PRIMARY       1 PrimaryHDU       4   ()
  1                1 BinTableHDU    250   22354100R x 121C   [K, 1A, K, K, 8A, K, 3A, D, D, D, D, D, D, D, D, D, D, D, D, D, D, D, D, D, D, D, D, D, D, D, D, D, D, D, D, D, D, D, K, K, K, K, K, K, K, D, D, D, D, D, D, D, D, D, D, D, D, D, D, D, D, D, D, D, D, K, K, K, K, K, K, K, K, D, D, D, D, D, D, D, D, D, D, D, 8A, D, D, D, D, D, D, D, D, D, D, D, D, 2A, K, D, D, D, D, D, D, D, D, K, D, D, K, D, D, D, D, D, D, K, K, D, D]

fitsinfo /global/cfs/cdirs/desi/target/catalogs/dr9/0.50.0/targets/sv1/secondary/dark/sv1targets-dark-secondary.fits
Filename: /global/cfs/cdirs/desi/target/catalogs/dr9/0.50.0/targets/sv1/secondary/dark/sv1targets-dark-secondary.fits
No.    Name      Ver    Type      Cards   Dimensions   Format
  0  PRIMARY       1 PrimaryHDU       6   ()
  1  SCND_TARGETS    1 BinTableHDU     71   22354100R x 16C   [D, D, E, E, E, L, E, E, E, K, K, K, D, K, K, K]
geordie666 commented 2 years ago

I can help review but I'm in meetings the rest of the day, so won't get to this until tomorrow morning.

geordie666 commented 2 years ago

I ran through 1000 secondary targets from each of the "bright" and "dark" files, and I confirm that everything seems correct in Raga's files. Here's the snippet of code I used:

rows_to_test = 1000 
import fitsio 
from time import time 
import numpy as np 
import os 
from desitarget.targets import encode_targetid 
for obscon in "bright", "dark": 
    secs = fitsio.read("/global/u1/r/raga19/data/sv1targets-{}-secondary-dr9photometry.fits".format(obscon), upper=True)    
    secsnophot = fitsio.read("/global/cfs/cdirs/desi/target/catalogs/dr9/0.50.0/targets/sv1/secondary/{}/sv1targets-{}-secondary.fits".format(obscon, obscon), upper=True)  
    secsnophot = secsnophot[secs["RA"] < 1e20]     
    secs = secs[secs["RA"] < 1e20] 
    print("Working with {} {} secondaries".format(len(secs), obscon)) 
    print("TARGETIDs match: {}".format(np.all(secs["TARGETID"] == secsnophot["TARGETID"]))) 
    alltrac = [] 
    start = time() 
    for i, objs in enumerate(secs[:rows_to_test]): 
        ns = ["south", "north"][objs["PHOTSYS"] =='N']  
        brickname = objs["BRICKNAME"] 
        fn = os.path.join("/global/cfs/cdirs/cosmo/data/legacysurvey/dr9/{}/tractor/".format(ns), "{}/tractor-{}.fits".format(brickname[:3], brickname)) 
        trac = fitsio.read(fn, upper=True) 
        alltrac.append(trac[trac["OBJID"] == objs["OBJID"]]) 
        if i%100 == 99: 
            print("Checked {}/{} secondaries...t={:.1f}s".format(i+1, len(secs), time()-start)) 
    alltrac = np.concatenate(alltrac) 
    tids = encode_targetid(objid=alltrac["OBJID"], brickid=alltrac["BRICKID"], release=alltrac["RELEASE"]) 
    print("Tractor TARGETIDs match: {} for {} secondaries".format(np.all(tids == secs[:rows_to_test]["TARGETID"]), rows_to_test))

Depending on how formal @moustakas wants to make these files, we might want more informative header information. We might also want a different schema than just setting all of the columns to "ridiculous" values when there is no match with the Legacy Surveys Tractor catalogs.

If these files are only ever used for spectrophotometric calibration of a few SV1 secondary targets, I'm not sure we want to be too pedantic, though.

moustakas commented 2 years ago

Thanks for the checks, @geordie666.

@Ragadeepika-Pucha, is it straightforward for you to make a couple changes to these catalogs? The biggest change is that the data type of each column in the catalogs you made should match the corresponding column in the sweep catalogs. (For example, in your catalogs sersic is float64 but in the sweeps it's float32. Also, in the sweeps the column names are capitalized, but that's a minor difference.)

I also agree with @geordie666 that for non-matches we should have simpler "blank" values. For non-matches it looks like you're using N/A for strings (although photsys for non-matches is N, which I think we should definitely change) and either 999999 or 1e+20 for floats. For simplicity, I suggest using zero (or minus one, although negative values can cause problems sometimes).

Are these changes easy for you?

geordie666 commented 2 years ago

In case it helps, the full data model for every column in the sweeps files is here:

https://github.com/legacysurvey/legacypipe/blob/6d1a92f8462f4db9360fb1a68ef7d6c252781027/bin/generate-sweep-files.py#L333-L547

Ragadeepika-Pucha commented 2 years ago

@moustakas -- I think it is pretty straightforward for me to 1) change the datatype of columns, and 2) change the column names to capital letters. However, I am not completely sure about how to insert zero values for all the columns (given different datatypes) for non-matches. I will give it a try and let you know. @geordie666 - Thank you for the datamodel.

Ragadeepika-Pucha commented 2 years ago

@moustakas - I have changed the datatype of columns and changed the names to capital letters. In the case of non-matches -- I filled it with 0 and 0.0 for int and float columns respectively -- and empty strings in 'S' dtype columns. There is a bool column - which is marked as '?' in dtype -- I filled the non-matches as False in this column. Is this fine? If yes, I will double-check everything and publish the new tables.

moustakas commented 2 years ago

@Ragadeepika-Pucha this all sounds great, thank you.

Ragadeepika-Pucha commented 2 years ago

The catalogs are sv1targets-bright-secondary-dr9photometry.fits and sv1targets-dark-secondary-dr9photometry.fits - These are located in the /global/u1/r/raga19/data/. To summarize -

Hopefully, this works fine. Let me know if there is anything more. @moustakas Thank you.

moustakas commented 2 years ago

Thank you, @Ragadeepika-Pucha! I spot-checked the new catalogs and everything looks good.

@geordie666 could you copy these catalogs into place so I can more fully test them within the fastspecfit pipeline? (I don't have permission to do so.)

cp /global/u1/r/raga19/data/sv1targets-dark-secondary-dr9photometry.fits /global/cfs/cdirs/desi/target/catalogs/dr9/0.50.0/targets/sv1/secondary/dark/
cp /global/u1/r/raga19/data/sv1targets-bright-secondary-dr9photometry.fits /global/cfs/cdirs/desi/target/catalogs/dr9/0.50.0/targets/sv1/secondary/bright/
geordie666 commented 2 years ago

OK, these files are now in the appropriate directories.

moustakas commented 2 years ago

Thanks, @geordie666.

I finally got around to testing these catalogs and realized that we missed a few other locations.

@Ragadeepika-Pucha, we need row-matched DR9 photometry for the following SV1 catalogs, too, in order for me to accommodate all the SV1 tiles. Is it straightforward for you to make them?

-r--r----- 1 62359 desi 257394240 Jan  9  2021 /global/cfs/cdirs/desi/target/catalogs/dr9/0.48.0/targets/sv1/secondary/bright/sv1targets-bright-secondary.fits
-r--r----- 1 62359 desi 214087680 Jan 19  2021 /global/cfs/cdirs/desi/target/catalogs/dr9/0.49.0/targets/sv1/secondary/bright/sv1targets-bright-secondary.fits
-r--r----- 1 62359 desi 252198720 Mar  7 15:07 /global/cfs/cdirs/desi/target/catalogs/dr9/0.51.0/targets/sv1/secondary/bright/sv1targets-bright-secondary.fits
-r--r----- 1 62359 desi 271535040 Mar 14 05:14 /global/cfs/cdirs/desi/target/catalogs/dr9/0.52.0/targets/sv1/secondary/bright/sv1targets-bright-secondary.fits

and

-r--r----- 1 62359 desi 4390352640 Jan  9  2021 /global/cfs/cdirs/desi/target/catalogs/dr9/0.48.0/targets/sv1/secondary/dark/sv1targets-dark-secondary.fits
-r--r----- 1 62359 desi 2177593920 Jan 19  2021 /global/cfs/cdirs/desi/target/catalogs/dr9/0.49.0/targets/sv1/secondary/dark/sv1targets-dark-secondary.fits
-r--r----- 1 62359 desi 2654876160 Mar  7 15:07 /global/cfs/cdirs/desi/target/catalogs/dr9/0.51.0/targets/sv1/secondary/dark/sv1targets-dark-secondary.fits
-r--r----- 1 62359 desi 2686466880 Mar 14 05:14 /global/cfs/cdirs/desi/target/catalogs/dr9/0.52.0/targets/sv1/secondary/dark/sv1targets-dark-secondary.fits
geordie666 commented 2 years ago

@moustakas: Are you sure that's the correct list? Anand showed me the following neat trick for determining which target versions were actually observed on tiles:

d = Table.read("/global/cfs/cdirs/desi/users/raichoor/fiberassign-rerun/fiberassign-history.ecsv")
d = d[d["SURVEY"] == "sv1"]
for cat in np.unique(d["DESITARGET_CAT"]):
    sel = d["DESITARGET_CAT"] == cat
    print("{} : {} tiles".format(cat, sel.sum()))
...
DESIROOT/target/catalogs/dr9/0.47.0/targets/cmx/resolve/no-obscon/ : 1 tiles
DESIROOT/target/catalogs/dr9/0.47.0/targets/sv1/resolve/bright/ : 47 tiles
DESIROOT/target/catalogs/dr9/0.47.0/targets/sv1/resolve/dark/ : 34 tiles
DESIROOT/target/catalogs/dr9/0.50.0/targets/sv1/resolve/bright : 18 tiles
DESIROOT/target/catalogs/dr9/0.50.0/targets/sv1/resolve/dark : 17 tiles
DESIROOT/target/catalogs/dr9/0.51.0/targets/sv1/resolve/dark : 29 tiles
DESIROOT/target/catalogs/gaiadr2/0.48.0/targets/sv1/resolve/supp : 2 tiles

So, it seems like we did observe version 0.47.0 for sv1 but we did not observe 0.52.0?

moustakas commented 2 years ago

What I did was much dumber, @geordie666. I didn't consider what was observed, I just did:

ls -l /global/cfs/cdirs/desi/target/catalogs/dr9/*/targets/sv1/secondary/bright/sv1targets-bright-secondary.fits
ls -l /global/cfs/cdirs/desi/target/catalogs/dr9/*/targets/sv1/secondary/dark/sv1targets-dark-secondary.fits

Maybe someone will want DR9 photometry for these intermediately-tagged catalogs even if we never got on-sky with them? I dunno.

In any case, @Ragadeepika-Pucha, in order for me to push your SV1 sample through fastspecfit I need DR9 photometry for tags 0.47.0 bright/dark and 0.51.0 dark (you already did 0.50.0 bright/dark and 0.48.0 is Gaia, which I don't think we care about.

geordie666 commented 2 years ago

Hey, Anand's the smart one. I think you're correct that we should just run all of them if @Ragadeepika-Pucha is willing? It's probably best to process the files while this is fresh-in-the-memory, even if nobody ultimately needs the extra catalogs.

Ragadeepika-Pucha commented 2 years ago

@moustakas @geordie666 - I will work on them this week and keep you posted. Thanks!

Ragadeepika-Pucha commented 2 years ago

@moustakas - Can you please confirm which files you are requiring photometry for?

0.47.0 does not have any secondary/* files. The sv1targets-*-secondary.fits files that I found are the following -

/global/cfs/cdirs/desi/target/catalogs/dr9/0.48.0/targets/sv1/secondary/bright/sv1targets-bright-secondary.fits /global/cfs/cdirs/desi/target/catalogs/dr9/0.49.0/targets/sv1/secondary/bright/sv1targets-bright-secondary.fits /global/cfs/cdirs/desi/target/catalogs/dr9/0.50.0/targets/sv1/secondary/bright/sv1targets-bright-secondary.fits /global/cfs/cdirs/desi/target/catalogs/dr9/0.51.0/targets/sv1/secondary/bright/sv1targets-bright-secondary.fits /global/cfs/cdirs/desi/target/catalogs/dr9/0.52.0/targets/sv1/secondary/bright/sv1targets-bright-secondary.fits

/global/cfs/cdirs/desi/target/catalogs/dr9/0.48.0/targets/sv1/secondary/dark/sv1targets-dark-secondary.fits /global/cfs/cdirs/desi/target/catalogs/dr9/0.49.0/targets/sv1/secondary/dark/sv1targets-dark-secondary.fits /global/cfs/cdirs/desi/target/catalogs/dr9/0.50.0/targets/sv1/secondary/dark/sv1targets-dark-secondary.fits /global/cfs/cdirs/desi/target/catalogs/dr9/0.51.0/targets/sv1/secondary/dark/sv1targets-dark-secondary.fits /global/cfs/cdirs/desi/target/catalogs/dr9/0.52.0/targets/sv1/secondary/dark/sv1targets-dark-secondary.fits

The resolve folder is probably not secondary targets? The 0.47.0 has only resolve folder. I am working on the 0.51.0 version now - Let me know if all the ten files mentioned above are the correct ones?

moustakas commented 2 years ago

That list looks right @Ragadeepika-Pucha! (I didn't mention 0.50.0 because you already made those catalogs.)

Ragadeepika-Pucha commented 2 years ago

@moustakas - all the files are ready and are located here - /global/cfs/cdirs/desi/users/raga19/data/ All the versions 0.*.0/ folders have one file corresponding to 'bright' and one file corresponding to 'dark' files, except 0.48.0.

The dark catalog for 0.48.0 had ~45.26 million rows and I couldn't save the whole table into a single fits file because of memory issues. The exclusive cpu node was not starting for me properly - so, I saved that into two tables - sv1targets_dark_secondary_p1.fits and sv1targets_dark_secondary_p2.fits. The first one has the first 22 million rows and the second one has the rest of the rows. I will try to combine them again tomorrow - but, my jupyter notebook keeps crashing.

Let me know if there are any other issues or anything else. Thank you.

moustakas commented 2 years ago

Thank you, @Ragadeepika-Pucha!

I was able to combine the troublesome catalogs in a login node with 

import fitsio
from astropy.table import Table, vstack
cd /global/cfs/cdirs/desi/users/raga19/data/0.48.0
t1 = Table(fitsio.read('sv1targets_dark_secondary_dr9_p1.fits'))
t2 = Table(fitsio.read('sv1targets_dark_secondary_dr9_p2.fits'))
out = vstack((t1, t2))
out.write('sv1targets_dark_secondary_dr9.fits')

And then I fixed permissions. I also spot-checked the merged catalog and everything looks OK as far as I can tell.

@geordie666 could we please copy these catalogs into place? Please feel free to double-check me on these commands! Also, I propose we overwrite the existing 0.50.0 DR9 photometric catalogs in case there were any changes to @Ragadeepika-Pucha's code.

cp /global/cfs/cdirs/desi/users/raga19/data/0.48.0/sv1targets_dark_secondary_dr9.fits /global/cfs/cdirs/desi/target/catalogs/dr9/0.48.0/targets/sv1/secondary/dark/sv1targets-dark-secondary-dr9photometry.fits 
cp /global/cfs/cdirs/desi/users/raga19/data/0.48.0/sv1targets_bright_secondary_dr9.fits /global/cfs/cdirs/desi/target/catalogs/dr9/0.48.0/targets/sv1/secondary/bright/sv1targets-bright-secondary-dr9photometry.fits 

cp /global/cfs/cdirs/desi/users/raga19/data/0.49.0/sv1targets_dark_secondary_dr9.fits /global/cfs/cdirs/desi/target/catalogs/dr9/0.49.0/targets/sv1/secondary/dark/sv1targets-dark-secondary-dr9photometry.fits 
cp /global/cfs/cdirs/desi/users/raga19/data/0.49.0/sv1targets_bright_secondary_dr9.fits /global/cfs/cdirs/desi/target/catalogs/dr9/0.49.0/targets/sv1/secondary/bright/sv1targets-bright-secondary-dr9photometry.fits 

cp /global/cfs/cdirs/desi/users/raga19/data/0.50.0/sv1targets_dark_secondary_dr9.fits /global/cfs/cdirs/desi/target/catalogs/dr9/0.50.0/targets/sv1/secondary/dark/sv1targets-dark-secondary-dr9photometry.fits 
cp /global/cfs/cdirs/desi/users/raga19/data/0.50.0/sv1targets_bright_secondary_dr9.fits /global/cfs/cdirs/desi/target/catalogs/dr9/0.50.0/targets/sv1/secondary/bright/sv1targets-bright-secondary-dr9photometry.fits 

cp /global/cfs/cdirs/desi/users/raga19/data/0.51.0/sv1targets_dark_secondary_dr9.fits /global/cfs/cdirs/desi/target/catalogs/dr9/0.51.0/targets/sv1/secondary/dark/sv1targets-dark-secondary-dr9photometry.fits 
cp /global/cfs/cdirs/desi/users/raga19/data/0.51.0/sv1targets_bright_secondary_dr9.fits /global/cfs/cdirs/desi/target/catalogs/dr9/0.51.0/targets/sv1/secondary/bright/sv1targets-bright-secondary-dr9photometry.fits 

cp /global/cfs/cdirs/desi/users/raga19/data/0.52.0/sv1targets_dark_secondary_dr9.fits /global/cfs/cdirs/desi/target/catalogs/dr9/0.52.0/targets/sv1/secondary/dark/sv1targets-dark-secondary-dr9photometry.fits 
cp /global/cfs/cdirs/desi/users/raga19/data/0.52.0/sv1targets_bright_secondary_dr9.fits /global/cfs/cdirs/desi/target/catalogs/dr9/0.52.0/targets/sv1/secondary/bright/sv1targets-bright-secondary-dr9photometry.fits
geordie666 commented 2 years ago

The catalogs have now been copied into place.

moustakas commented 2 years ago

Thanks @geordie666.

Testing a handful of @Ragadeepika-Pucha's galaxies with

fastspec /global/cfs/cdirs/desi/spectro/redux/everest/healpix/sv1/dark/91/9157/redrock-sv1-dark-9157.fits \
  -o fastspec.fits --specprod everest --mp 7 \
  --targetids 39632936277902799,39632931181824699,39632936277905113,39632936273709365,39632936282096442,39632936273708359,39632931177631304

I get an error:

INFO:io.py:166:find_specfiles: Parsed coadd_type=healpix
INFO:io.py:63:_get_targetdirs: Reading /global/cfs/cdirs/desi/target/fiberassign/tiles/trunk/080/fiberassign-080896.fits.gz header.
INFO:io.py:341:find_specfiles: Matched 3 targets in /global/cfs/cdirs/desi/target/catalogs/dr9/0.51.0/targets/sv1/resolve/dark/sv1targets-dark-hp-143.fits
INFO:io.py:341:find_specfiles: Matched 4 targets in /global/cfs/cdirs/desi/target/catalogs/dr9/0.51.0/targets/sv1/secondary/dark/sv1targets-dark-secondary-dr9photometry.fits
INFO:io.py:63:_get_targetdirs: Reading /global/cfs/cdirs/desi/target/fiberassign/tiles/trunk/080/fiberassign-080895.fits.gz header.
INFO:io.py:341:find_specfiles: Matched 3 targets in /global/cfs/cdirs/desi/target/catalogs/dr9/0.51.0/targets/sv1/resolve/dark/sv1targets-dark-hp-143.fits
INFO:io.py:341:find_specfiles: Matched 4 targets in /global/cfs/cdirs/desi/target/catalogs/dr9/0.51.0/targets/sv1/secondary/dark/sv1targets-dark-secondary-dr9photometry.fits
Traceback (most recent call last):
  File "/global/homes/i/ioannis/repos/desihub/fastspecfit/bin/fastspec", line 9, in <module>
    fastspec(comm=None)
  File "/global/homes/i/ioannis/repos/desihub/fastspecfit/py/fastspecfit/fastspecfit.py", line 204, in fastspec
    Spec.find_specfiles(args.redrockfiles, firsttarget=args.firsttarget,
  File "/global/homes/i/ioannis/repos/desihub/fastspecfit/py/fastspecfit/io.py", line 346, in find_specfiles
    targets = Table(np.hstack(targets))
  File "<__array_function__ internals>", line 5, in hstack
  File "/opt/conda/lib/python3.8/site-packages/numpy/core/shape_base.py", line 344, in hstack
    return _nx.concatenate(arrs, 0)
  File "<__array_function__ internals>", line 5, in concatenate
ValueError: could not convert string to float: 'N'

If I try to use astropy.table.vstack I get

*** NotImplementedError: vstack unavailable for mixin column type(s): NdarrayMixin

so something is amiss.

Unfortunately I'm out of time right now to dig in case anyone else is inspired...

Ragadeepika-Pucha commented 2 years ago

@moustakas - The error comes from the PHOTSYS column from the catalogs I created. The datatype of TARGETID matches with the targets catalog and the datatype of LS DR9 columns match the sweep files. However, the PHOTSYS column is 'S1' in my catalogs, while it is '<U1' in all the everest files. I am sorry for the confusion - I am making the changes on all the catalogs now. @geordie666 - Is it possible to make the changes directly to the catalogs wherever you have staged them? Or should I let you know about the changed catalogs?

geordie666 commented 2 years ago

@Ragadeepika-Pucha: Let me know the location of the changed files once they're ready and then I'll transfer them to the write-protected locations.

Ragadeepika-Pucha commented 2 years ago

@moustakas - I finished making the changes to the PHOTSYS column for all the catalogs. The 0.48.0 catalogs - *p1.fits and *p2.fits needs to be combined again. Can you test the code again using these versions before transferring them to their locations? Thanks!

moustakas commented 2 years ago

Thanks @Ragadeepika-Pucha, I'll test the catalogs but if you could merge your two partial catalogs that'd be helpful. I gave the snippet of code above and you just need to log in and do

source /global/cfs/cdirs/desi/software/desi_environment.sh master
Ragadeepika-Pucha commented 2 years ago

@moustakas - I cannot merge the two partial catalogs because of memory issues. I am uable to use my exlusive cpu node access and the normal cori access keeps crashing because of the huge number of rows.

moustakas commented 2 years ago

I'm getting the same error. I don't quite understand it, but it looks like it's the order of the columns in your catalog that's the issue, and so when I try to stack them, np.hstack complains.

Here's a snippet of output just so I don't forget (no action items, I'm working on it).

(Pdb) Table(fitsio.read('/global/cfs/cdirs/desi/target/catalogs/dr9/0.51.0/targets/sv1/resolve/dark/sv1targets-dark-hp-143.fits', rows=0, columns=targetcols))
<Table length=1>
       RA               DEC          FLUX_G    FLUX_R    FLUX_Z  FLUX_IVAR_G ... FIBERFLUX_Z FIBERTOTFLUX_G FIBERTOTFLUX_R FIBERTOTFLUX_Z PHOTSYS      TARGETID
    float64           float64       float32   float32   float32    float32   ...   float32      float32        float32        float32       str1        int64
---------------- ----------------- --------- --------- --------- ----------- ... ----------- -------------- -------------- -------------- ------- -----------------
227.906196819581 32.89192548503159 5.0230317 18.375298 54.065174    372.7904 ...    41.97429       3.901896      14.273912       42.05138       N 39632941373980741
(Pdb) Table(targets[0])
<Table length=3>
        RA                DEC          FLUX_G    FLUX_R    FLUX_Z   FLUX_IVAR_G ... FIBERFLUX_Z FIBERTOTFLUX_G FIBERTOTFLUX_R FIBERTOTFLUX_Z PHOTSYS      TARGETID
     float64            float64       float32   float32   float32     float32   ...   float32      float32        float32        float32       str1        int64
------------------ ----------------- --------- --------- ---------- ----------- ... ----------- -------------- -------------- -------------- ------- -----------------
226.96923847156185 32.44379674657387 1.6316719 3.2548742  2.3007507    320.8553 ...   1.7891678      1.2688626      2.5311382      1.7891685       N 39632931177631304
 227.2240108136138 32.68367592661154 4.2744813  8.934247  14.078902   256.47095 ...    5.350051       1.625701      3.3975642      5.3531094       N 39632936282096442
227.14437016623828 32.75792289396391  4.084077 9.0138445 15.2085285   427.11798 ...    8.869176      2.3817163      5.2566147       8.869176       N 39632936277905113
(Pdb) Table(targets[1])
<Table length=4>
     TARGETID     PHOTSYS         RA                DEC          FLUX_G   FLUX_R  ... FIBERFLUX_G FIBERFLUX_R FIBERFLUX_Z FIBERTOTFLUX_G FIBERTOTFLUX_R FIBERTOTFLUX_Z
      int64         str1       float64            float64       float32  float32  ...   float32     float32     float32      float32        float32        float32
----------------- ------- ------------------ ------------------ -------- -------- ... ----------- ----------- ----------- -------------- -------------- --------------
39632931181824699       N  227.2130001384101 32.414673886647066 1.594635 4.088936 ...    0.908661    2.329973    4.562993       0.908661       2.329973       4.562993
39632936277902799       N 226.97106504521375 32.825430731213345 2.542069 6.002813 ...    1.126328    2.659698    4.918194       1.126328       2.659698       4.918194
39632936273708359       N 226.66432745704736  32.67825968016942 2.408377 4.597582 ...    1.344683    2.566995     4.09719       1.344683       2.566995        4.09719
39632936273709365       N  226.7409700725095  32.81154396638427 2.066544 6.093318 ...    1.300753     3.83534    7.052577       1.300753        3.83534       7.052577
Ragadeepika-Pucha commented 2 years ago

If it is the order of the columns -- then you can maybe try this -

If Table(targets[0]) is table1 and Table(targets[1]) is table2 then - columns = table1.colnames table2_new = table2[columns] stacked_table = astropy.table.vstack((table1, table2_new))

I am not sure why you want to do hstack - if the purpose is to make one single table with all the sources, then shouldn't we be using vstack?

moustakas commented 2 years ago

OK, I capture the outstanding issue in code in fastspecfit but if we think anyone else is going to be stacking these SV1 catalogs with non-SV1 catalogs then we should probably fix this.

@geordie666 can you please move the files into their final resting place? Note that I combined the 0.48.0 dark catalogs for @Ragadeepika-Pucha as requested, so we should be good.

@Ragadeepika-Pucha, using np.hstack with a list of np.ndarray tables is much faster than using astropy.table.vstack on a list of astropy.table.Table tables. I can't remember who I learned this trick from---I think @tskisner---but whoever it was, thank you!

geordie666 commented 2 years ago

OK, files should now be in place with the correct permissions at:

ls -alrt /global/cfs/cdirs/desi/target/catalogs/dr9/*/targets/sv1/secondary/*/sv1targets-*-secondary-dr9photometry.fits
moustakas commented 2 years ago

Thanks @geordie666. If there's justice in the world, this will close this ticket. (But I'll reopen if there are any issues in production...)

moustakas commented 2 years ago

Unfortunately it looks like the main-survey secondary target catalogs are also missing PHOTSYS, among other quantities:

fitsio.FITS(targetfile)['SCND_TARGETS']

  file: /global/cfs/cdirs/desi/target/catalogs/dr9/1.1.1/targets/main/secondary/dark/targets-dark-secondary.fits
  extension: 1
  type: BINARY_TBL
  extname: SCND_TARGETS
  rows: 7125595
  column info:
    RA                  f8
    DEC                 f8
    PMRA                f4
    PMDEC               f4
    REF_EPOCH           f4
    OVERRIDE            b1
    FLUX_G              f4
    FLUX_R              f4
    FLUX_Z              f4
    PARALLAX            f4
    GAIA_PHOT_G_MEAN_MAG
                        f4
    GAIA_PHOT_BP_MEAN_MAG
                        f4
    GAIA_PHOT_RP_MEAN_MAG
                        f4
    GAIA_ASTROMETRIC_EXCESS_NOISE
                        f4
    TARGETID            i8
    DESI_TARGET         i8
    SCND_TARGET         i8
    SCND_ORDER          i4
    SUBPRIORITY         f8
    OBSCONDITIONS       i8
    PRIORITY_INIT       i8
    NUMOBS_INIT         i8
geordie666 commented 2 years ago

Because (at least, as I recall) I always had the code in place for matching back to the Legacy Surveys for Main Survey secondary targets, I think that there should be two differences with the Main Survey files:

(1) The fluxes should be included. (2) The incarnation at the top of this thread should recover PHOTSYS.

I'm not sure if you want other columns (e.g. the WISE fluxes) but (again, I think) at least (1) and (2) should be available.

moustakas commented 2 years ago

I think this issue has been handled in the latest (Everest) release, but let's open a new ticket if other issues are uncovered.