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
When a <Sample> specifies the HistoPath and it's not empty, we will run into very frequent deserialization errors from uproot which is very slow to raise an exception (see scikit-hep/uproot5#504).
pyhf needs to fix this to be a bit smarter in how to check valid keys, and in particular, fix up its logic to not hit a/rely on DeserializationError.
OS / Environment
$ system_profiler -detailLevel mini SPSoftwareDataType | head -n 6Software: System Software Overview: System Version: macOS 10.14.6 (18G9323) Kernel Version: Darwin 18.7.0
Steps to Reproduce
See #1687 for the fundamental issue. It's reproducible using private workspaces for now, but can be confirmed reproducible.
$ time pyhf xml2json monotop_twmetComb0L1LBoosted_allCRs_normDStoDR_unblind_sigTheo_envelope__pmoder_sig_a250_DM10_H900_tb1_st0p7/NormalMeasurement.xml --basedir monotop_twmetComb0L1LBoosted_allCRs_normDStoDR_unblind_sigTheo_envelope__pmoder_sig_a250_DM10_H900_tb1_st0p7
Processing ./NormalMeasurement_CR_0LBoosted_ttbar_cuts.xml: 0%| | 0/1 [00:00<?, ?channel/smonotop_twmetComb0L1LBoosted_allCRs_normDStoDR_unblind_sigTheo_envelope__pmoder_sig_a250_DM10_H900_tb1_st0p7/results/monotop_twmetComb0L1LBoosted_allCRs_normDStoDR_unblind_sigTheo_envelope__pmoder_sig_a250_DM10_H900_tb1_st0p7/Exclusion_combined_NormalMeasurement_model.root not in filecache
path=CR_0LBoosted_ttbar_cuts_hists/data, name=hData_CR_0LBoosted_ttbar_obs_cuts
deserialization error, trying fullname=CR_0LBoosted_ttbar_cuts_hists/data/hData_CR_0LBoosted_ttbar_obs_cuts instead
monotop_twmetComb0L1LBoosted_allCRs_normDStoDR_unblind_sigTheo_envelope__pmoder_sig_a250_DM10_H900_tb1_st0p7/results/monotop_twmetComb0L1LBoosted_allCRs_normDStoDR_unblind_sigTheo_envelope__pmoder_sig_a250_DM10_H900_tb1_st0p7/Exclusion_combined_NormalMeasurement_model.root in filecache
path=CR_0LBoosted_ttbar_cuts_hists/Top0LBoosted, name=hTop0LBoostedNom_CR_0LBoosted_ttbar_obs_cuts
deserialization error, trying fullname=CR_0LBoosted_ttbar_cuts_hists/Top0LBoosted/hTop0LBoostedNom_CR_0LBoosted_ttbar_obs_cuts instead
monotop_twmetComb0L1LBoosted_allCRs_normDStoDR_unblind_sigTheo_envelope__pmoder_sig_a250_DM10_H900_tb1_st0p7/results/monotop_twmetComb0L1LBoosted_allCRs_normDStoDR_unblind_sigTheo_envelope__pmoder_sig_a250_DM10_H900_tb1_st0p7/Exclusion_combined_NormalMeasurement_model.root in filecache
path=CR_0LBoosted_ttbar_cuts_hists/Top0LBoosted, name=hTop0LBoostedEG_EffLow_CR_0LBoosted_ttbar_obs_cutsNorm
deserialization error, trying fullname=CR_0LBoosted_ttbar_cuts_hists/Top0LBoosted/hTop0LBoostedEG_EffLow_CR_0LBoosted_ttbar_obs_cutsNorm instead
File Upload (optional)
No response
Expected Results
pyhf xml2json should be fast.
Actual Results
`pyhf xml2json` is slow.
pyhf Version
This impacts all pyhf versions up to 0.6.4.
Code of Conduct
I agree to follow the Code of Conduct
The text was updated successfully, but these errors were encountered:
Summary
When a
<Sample>
specifies theHistoPath
and it's not empty, we will run into very frequent deserialization errors fromuproot
which is very slow to raise an exception (see scikit-hep/uproot5#504).pyhf
needs to fix this to be a bit smarter in how to check valid keys, and in particular, fix up its logic to not hit a/rely onDeserializationError
.OS / Environment
Steps to Reproduce
See #1687 for the fundamental issue. It's reproducible using private workspaces for now, but can be confirmed reproducible.
File Upload (optional)
No response
Expected Results
pyhf xml2json
should be fast.Actual Results
`pyhf xml2json` is slow.
pyhf Version
This impacts all pyhf versions up to 0.6.4.
Code of Conduct
The text was updated successfully, but these errors were encountered: