Remove eostools dependency from cmsIO #184
Conversation
|
Ciao Giovanni, many thanks! It works, but the file used by heppy_batch is the same, but in PhysicsTools: |
|
@GaelTouquet tested it for H2TauTau, and it also appears to work fine in our case. Thanks for the quick fix!
… On 13 Jun 2017, at 14:26, Giovanni Petrucciani ***@***.***> wrote:
cmsIO used to be taken from the CAF AFS installation, but it is now deprecated.
This small change to eostools avoids using cmsIO, and in the test case I've run things seem to work fine, but a few more tests would be welcome (especially if other analysis setups make a more complex use of eostools wrt the multilepton cfg I tested)
Many eostools are still littered with multiple translations from lfn to pfn and back again that to me don't seem necessary (maybe they were introduced to make sure the paths had a canonical form?) but I didn't remove them yet.
There's probably also a lot of obsolete legacy API (e.g for castor) that could be cleaned up.
A volunteer would be welcome.
You can view, comment on, or merge this pull request online at:
#184
Commit Summary
• eostools: avoid cmsIO module that is no longer available in the standard CAF environment
• samples_13TeV_80X_susySignalsPriv: Remove missing private Higgsino samples
File Changes
• M Production/python/eostools.py (145)
• M RootTools/python/samples/samples_13TeV_80X_susySignalsPriv.py (10)
Patch Links:
• https://github.com/CERN-PH-CMG/cmgtools-lite/pull/184.patch
• https://github.com/CERN-PH-CMG/cmgtools-lite/pull/184.diff
—
You are receiving this because you are subscribed to this thread.
Reply to this email directly, view it on GitHub, or mute the thread.
|
cmsIO fix (CERN-PH-CMG/cmgtools-lite#184) also in HeppyCore
|
|
||
| def splitPFN(pfn): | ||
| """Split the PFN in to { <protocol>, <host>, <path>, <opaque> }""" | ||
| groups = re.match("^(\w\+)://([^/]+)/(/[^?]+)(\?.*)?", pfn) |
There was a problem hiding this comment.
I think I missed one problem here, the backslash before the first + should not be there. Instead, it should be:
groups = re.match("^(\w+)://([^/]+)/(/[^?]+)(\?.*)?", pfn)
which for instance works for root://eoscms.cern.ch//eos/cms/store/cmst3/user/jngadiub/ntuple returning
root
eoscms.cern.ch
/eos/cms/store/cmst3/user/jngadiub/ntuple
| lfn = eosToLFN(path) | ||
| pfn = lfnToPFN(lfn) | ||
| tokens = cmsIO.splitPFN(pfn) | ||
| tokens = splitPFN(path) |
There was a problem hiding this comment.
There is another issue here (or in general), it calls splitPFN with a /eos/... path and not with the protocol prepended, so the regex evaluation above will fail.
There was a problem hiding this comment.
I think the error is later on at lines 194-196. It works when replacing those lines with:
pfn = lfnToPFN(path)
if not isEOSFile(pfn):
runEOSCommand(pfn,'mkdir','-p')
typo in filename
cmsIO used to be taken from the CAF AFS installation, but it is now deprecated.
This small change to eostools avoids using cmsIO, and in the test case I've run things seem to work fine, but a few more tests would be welcome (especially if other analysis setups make a more complex use of eostools wrt the multilepton cfg I tested)
Many eostools are still littered with multiple translations from lfn to pfn and back again that to me don't seem necessary (maybe they were introduced to make sure the paths had a canonical form?) but I didn't remove them yet.
There's probably also a lot of obsolete legacy API (e.g for castor) that could be cleaned up.
A volunteer would be welcome.