You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Open question: what do you want to do with HSC and DES?
Just delete: drop the stile import, remove the HSC and DES branches + the Rho{1..5}SysTest subclasses, switch the 5 HSC example configs to UNIONS.
Migrate: re-route HSC and DES through shear_psf_leakage so the existing RHO_STATS_STYLE knob keeps working, just without stile.
Delete-and-go is simplest. Migrate preserves compatibility but needs a numerical sanity check. Happy to go either way — and if there's still real-data production relying on HSC-style rho-stats, that probably pushes toward migrate.
I opt for 1, delete. In fact, I believe that the rho stats within shapepipe are obsolete, and have been superceded by sp_validation / cosmo_val, which uses shear_psf_leakage.
Resolved by #715 (merged 2026-04-23), which removed stile and the rho-stats code that depended on it. Install should no longer hit the Stile fork. The residual treecorr docs cleanup is in #716.
Activity
martinkilbinger commentedon Apr 22, 2026
New suggestions is to remove stile altogether and adapt the code accordingly. See #710 .
cailmdaley commentedon Apr 22, 2026
Following up on your suggestion to remove stile — here's what it touches and the one open question.
All uses live in
src/shapepipe/modules/mccd_package/mccd_plot_utilities.py(+mccd_plots_runner.py). Three rho-stat backends:RHO_STATS_STYLEHSCstile.CorrelationFunctionSysTestDESRho{1..5}SysTestsubclasses (~500 lines)UNIONSshear_psf_leakage.rho_tau_stat.RhoStatExample config usage:
HSC: 5 configsUNIONS: 2 (config_Pl_psfex.ini,config_Pl_psfex_conv.ini)DES: 0Open question: what do you want to do with HSC and DES?
HSCandDESbranches + theRho{1..5}SysTestsubclasses, switch the 5 HSC example configs toUNIONS.HSCandDESthroughshear_psf_leakageso the existingRHO_STATS_STYLEknob keeps working, just without stile.Delete-and-go is simplest. Migrate preserves compatibility but needs a numerical sanity check. Happy to go either way — and if there's still real-data production relying on HSC-style rho-stats, that probably pushes toward migrate.
martinkilbinger commentedon Apr 22, 2026
I opt for 1, delete. In fact, I believe that the rho stats within shapepipe are obsolete, and have been superceded by sp_validation / cosmo_val, which uses shear_psf_leakage.
cailmdaley commentedon Apr 24, 2026
Resolved by #715 (merged 2026-04-23), which removed stile and the rho-stats code that depended on it. Install should no longer hit the Stile fork. The residual treecorr docs cleanup is in #716.
refactor: remove rho-stats from shapepipe