source: TI01-discovery/trunk/ingestAutomation/OAIBatch/oai_ingest.py @ 1755

#!/usr/bin/env python
""" Script oai_ingest.py takes parameter <datacentre>.
The /usr/local/WSClients/OAIBatch directory contains this python script, a config file
and the oaiClean.py class which cleans up discovery records after harvesting.
The pre-processed files are then ingested to the eXist XML db.

 Under this directory the following structure should be maintained:

 ./data
 - /DATACENTRE/
                - discovery/:         Records with namespace, schema declaration deleted - after having run
                                      the oaiClean script. Ready to ingest in the discovery service.
                - oai/difYYYYMMDD/    Records as harvested from OAI

 Where  /DATACENTRE  varies for the different data providers

"""
#History:
# 12/05/06 SEL spelling correction
# 30/05/06 SEL cope with many files for processing."Argument list too long" problem.
# 31/05/06 SEL need to use passwords now. (replace xxxxxx in real version).
# 16/10/06 SEL Changed to using python oaiClean.py module instead of java code.
# 16/10/06 SEL exist db upgraded and deployed at different location, java upgrade.
# 17/10/06 SEL cope with different discovery formats - not just DIF.
# 23/10/06 SEL keywords not mandatory in config file.
# 24/10/06 SEL fix bug where 'in' directory not being cleared initially. display more information when running.

import os
import sys
import commands
import string
import oaiClean

status = 0
numfilesproc = 0
harvest_home = ""
datacentre_groups = ""
datacentre_format = ""

if (len(sys.argv) < 2):
    print "<datacentre>  parameter not supplied."
    sys.exit()
else:
    datacentre = sys.argv[1]

# Other settings and constants
date_string = commands.getoutput ("date +'%y%m%d_%H%M'")
os.putenv ('EXIST_HOME', '/usr/local/exist-client')
os.putenv ('PATH', ':/usr/java/jdk1.5.0_03/jre:/usr/java/jdk1.5.0_03:/usr/java/jdk1.5.0_03/lib/tools.jar:/usr/local/WSClients/OAIBatch:/usr/local/exist-client/bin:/bin:/usr/bin:.')
os.putenv ('CLASSPATH','.:/usr/java/j2sdk1.4.2_04/bin:/usr/local/WSClients/OAIBatch')

# Get the harvested records directory and groups for this datacentre from the datacentre specific config file
# The harvested records directory depends on the datacentres OAI base url, the set and  format. These have to be know up-front.
# The groups denote which 'portal groups' they belong to - for limiting searches to say NERC-only datacentres records.
# Groups are added to the xml record by oaiClean.py.
datacentre_config_filename = "/usr/local/WSClients/OAIBatch/" + datacentre + "_config.properties"
print "Datacentre config file = %s" %datacentre_config_filename
datacentre_config_file = open(datacentre_config_filename, "r")

for line in datacentre_config_file.readlines():
    words  = string.split(line)
    if len(words) == 0:
        continue
    if words[0] == 'host_path':
        harvest_home = string.rstrip(words[1])
    if words[0] == 'groups':
        datacentre_groups = words[1:]
    if words[0] == 'format':
        datacentre_format = words[1]
datacentre_config_file.close()
if harvest_home == "":
    sys.exit("Failed at getting harvested records directory stage. datacentre config file tried = %s" %datacentre_config_filename)
else:
    print "INFO: harvested records are in %s" %harvest_home

if datacentre_groups == "":
    print "INFO: No groups/keywords set for datacentre %s" %datacentre
else:
    print "INFO: datacentre groups/keywords = %s" %datacentre_groups

if datacentre_format == "":
    sys.exit("Failed at stage: getting datacentre format. datacentre config file tried = %s" %datacentre_config_filename)
else:
    print "INFO: format being harvested = %s" %datacentre_format

#any records to harvest?
if len( os.listdir(harvest_home)) == 0:
    print "Nothing to harvest this time from %s" %datacentre
    sys.exit()

# The directory to put things for a tape backup (should already exist)
backupdir = '/disks/glue1/oaiBackup/'

# Create/clear the 'in' directory pristine copy of the discovery records
if os.path.isdir("/usr/local/WSClients/OAIBatch/data/" + datacentre +"/oai/originals"):
    commandline = "ls -1 /usr/local/WSClients/OAIBatch/data/" + datacentre +"/oai/originals/ | xargs -i rm /usr/local/WSClients/OAIBatch/data/" + datacentre +"/oai/originals/{\}"
    print "Executing : " + commandline
    status = os.system(commandline)
else:
    commandline = "mkdir /usr/local/WSClients/OAIBatch/data/" + datacentre +"/oai/originals"
    print "Executing : " + commandline
    status= os.system(commandline)

if status != 0:
    sys.exit("Failed at creating copy dir stage")

# make the 'in' pristine copy. Cope with there being lots of files in the directory.

commandline = "ls -1 " + harvest_home + "/ | xargs -i cp " + harvest_home + "/{\} /usr/local/WSClients/OAIBatch/data/" + datacentre + "/oai/originals"
print "Executing : " + commandline
status = os.system(commandline)
if status !=0:
    sys.exit("Failed at making pristine copy stage")

# Create/clear the directory for the 'out' processed copy of the discovery records.
if os.path.isdir("/usr/local/WSClients/OAIBatch/data/" + datacentre +"/discovery"):
    commandline = "ls -1 /usr/local/WSClients/OAIBatch/data/" + datacentre +"/discovery/ | xargs -i rm /usr/local/WSClients/OAIBatch/data/" + datacentre +"/discovery/{\}"
    print "Executing : " + commandline
    status = os.system(commandline)
else:
    commandline = "mkdir /usr/local/WSClients/OAIBatch/data/" + datacentre +"/discovery"
    print "Executing : " + commandline
    status= os.system(commandline)

# Removed 16/10/06 - directory will hold 'out' processed records only
# make the processing copy
#commandline = "ls -1 " + harvest_home + "/ | xargs -i cp " + harvest_home + "/{\} /usr/local/WSClients/OAIBatch/data/" + datacentre + "/discovery"
#print "Executing : " + commandline
#status = os.system(commandline)
#if status !=0:
#    sys.exit("Failed at making processing copy stage")

# The file config.properties contains the location of the particular datacentres harvested records.
# Copy the datacentre specific version of config to config.properties file.
commandline = "cp /usr/local/WSClients/OAIBatch/" + datacentre +"_config.properties /usr/local/WSClients/OAIBatch/config.properties"
print "Executing : " + commandline
status = os.system(commandline)
if status !=0:
    sys.exit("Failed at copying config file stage")

#Change os directory to that with the oaiClean.py in it. (need this?)
os.chdir('/usr/local/WSClients/OAIBatch')


#Execute the script which processes the files
indir="/usr/local/WSClients/OAIBatch/data/" + datacentre +"/oai/originals"
outdir="/usr/local/WSClients/OAIBatch/data/" + datacentre +"/discovery"
wrapFlag=False
filenames = os.listdir(indir)
for filename in filenames:
        if filename.find('.xml') != -1:
                full_filename = indir + "/" + filename
                print "Processing : " + full_filename
                #try:
                oaiClean.oaiClean(indir,outdir,filename,wrapFlag)
                #except:
                #    sys.exit("Failed at processing file %s with oaiClean.py stage with status %s" %(full_filename, sys.exc_info()))
                #    break
                numfilesproc += 1
        else:
                print 'File %s is not xml format. Not processed'  %(full_filename)

# Removed 16/10/06 Don't need this anymore
#Once the pre-processing has finished remove the originals from the discovery directory:
#commandline = "find /usr/local/WSClients/OAIBatch/data/" + datacentre +"/discovery/oai* -print | xargs -i rm /{\}"
#print "Executing : " + commandline
#status = os.system(commandline)
#if status !=0:
#    sys.exit("Failed at removing original oai style records from discovery directory")


# ingest the datacentres records into eXist db (backups of exist happen nightly).
commandline = "$EXIST_HOME/bin/client.sh -c /db/discovery/" +datacentre_format+ "/" + datacentre + " -u admin -P xxxxxx -p /usr/local/WSClients/OAIBatch/data/" + datacentre +"/discovery/"
print "Executing : actual command to ingest into exist db"
status = os.system(commandline)
if status !=0:
    sys.exit("Failed at ingesting into exist db. Datacentre =  %s. Status = %s" %(datacentre,status))

#Make copies of discovery and oai/originals areas to backup area for tape backups
this_backupdir = backupdir + datacentre + "_" + date_string + "_originals"
commandline = "mkdir " + this_backupdir
print "Executing : " + commandline
status = os.system(commandline)
if status !=0:
    sys.exit("Failed at creating backup directory %s" %this_backupdir)

commandline = "ls -1 /usr/local/WSClients/OAIBatch/data/" + datacentre +"/oai/originals/ | xargs -i cp /usr/local/WSClients/OAIBatch/data/" + datacentre +"/oai/originals/{\} " + this_backupdir
print "Executing : " + commandline
status = os.system(commandline)
if status !=0:
    sys.exit("Failed at copying to backup directory %s" %this_backupdir)

this_backupdir = backupdir + datacentre + "_" + date_string + "_discovery"
commandline = "mkdir " + this_backupdir
print "Executing : " + commandline
status = os.system(commandline)
if status !=0:
    sys.exit("Failed at creating backup directory %s" %this_backupdir)

commandline = "ls -1 /usr/local/WSClients/OAIBatch/data/" + datacentre +"/discovery/ | xargs -i cp /usr/local/WSClients/OAIBatch/data/" + datacentre +"/discovery/{\} " + this_backupdir
print "Executing : " + commandline
status = os.system(commandline)
if status !=0:
    sys.exit("Failed at copying to backup directory %s" %this_backupdir)

#Clear out the original harvest records area
commandline = "ls -1 " + harvest_home + " | xargs -i rm " + harvest_home + "/{\}"
print "Executing : " + commandline
status = os.system(commandline)
if status !=0:
    sys.exit("Failed at clearing out original harvest records area %s" %harvest_home)


print "======================================================"
print "No. of files pre-processed = %s" %numfilesproc
if status == 0:
    print " Procedure oai_ingest.py ran to end"
else:
    print "Procedure oai_ingest.py FAILED with status %s" %status

print "======================================================"