def from_disk(self, resource_list=None, paths=None):
"""Create or extend resource_list with resources from disk scan
Assumes very simple disk path to URL mapping (in self.mapping): chop
path and replace with url_path. Returns the new or extended ResourceList
object.
If a resource_list is specified then items are added to that rather
than creating a new one.
If paths is specified then these are used instead of the set
of local paths in self.mapping.
Example usage with mapping start paths:
mapper=Mapper('http://example.org/path','/path/to/files')
rlb = ResourceListBuilder(mapper=mapper)
m = rlb.from_disk()
Example usage with explicit paths:
mapper=Mapper('http://example.org/path','/path/to/files')
rlb = ResourceListBuilder(mapper=mapper)
m = rlb.from_disk(paths=['/path/to/files/a','/path/to/files/b'])
"""
num = 0
# Either use resource_list passed in or make a new one
if (resource_list is None):
resource_list = ResourceList()
# Compile exclude pattern matches
self.compile_excludes()
# Work out start paths from map if not explicitly specified
if (paths is None):
paths = []
for map in self.mapper.mappings:
paths.append(map.dst_path)
# Set start time unless already set (perhaps because building in chunks)
if (resource_list.md_at is None):
resource_list.md_at = datetime_to_str()
# Run for each map in the mappings
for path in paths:
self.logger.info("Scanning disk from %s" % (path))
self.from_disk_add_path(path=path, resource_list=resource_list)
# Set end time
resource_list.md_completed = datetime_to_str()
return (resource_list)
def from_disk(self, resource_list=None, paths=None):
"""Create or extend resource_list with resources from disk scan
Assumes very simple disk path to URL mapping (in self.mapping): chop
path and replace with url_path. Returns the new or extended ResourceList
object.
If a resource_list is specified then items are added to that rather
than creating a new one.
If paths is specified then these are used instead of the set
of local paths in self.mapping.
Example usage with mapping start paths:
mapper=Mapper('http://example.org/path','/path/to/files')
rlb = ResourceListBuilder(mapper=mapper)
m = rlb.from_disk()
Example usage with explicit paths:
mapper=Mapper('http://example.org/path','/path/to/files')
rlb = ResourceListBuilder(mapper=mapper)
m = rlb.from_disk(paths=['/path/to/files/a','/path/to/files/b'])
"""
num=0
# Either use resource_list passed in or make a new one
if (resource_list is None):
resource_list = ResourceList()
# Compile exclude pattern matches
self.compile_excludes()
# Work out start paths from map if not explicitly specified
if (paths is None):
paths=[]
for map in self.mapper.mappings:
paths.append(map.dst_path)
# Set start time unless already set (perhaps because building in chunks)
if (resource_list.md_at is None):
resource_list.md_at = datetime_to_str()
# Run for each map in the mappings
for path in paths:
self.logger.info("Scanning disk from %s" % (path))
self.from_disk_add_path(path=path, resource_list=resource_list)
# Set end time
resource_list.md_completed = datetime_to_str()
return(resource_list)
def _str_datetime_now(self, x=None):
"""Return datetime string for use with time attributes
Handling depends on input:
'now' - returns datetime for now
number - assume datetime values, generate string
other - no change, return same value
"""
if (x == 'now'):
# Now, this is wht datetime_to_str() with no arg gives
return( datetime_to_str() )
try:
# Test for number
junk = x + 0.0
return datetime_to_str(x)
except TypeError:
# Didn't look like a number, treat as string
return x
def _str_datetime_now(self, x=None):
"""Return datetime string for use with time attributes
Handling depends on input:
'now' - returns datetime for now
number - assume datetime values, generate string
other - no change, return same value
"""
if (x == 'now'):
# Now, this is wht datetime_to_str() with no arg gives
return (datetime_to_str())
try:
# Test for number
junk = x + 0.0
return datetime_to_str(x)
except TypeError:
# Didn't look like a number, treat as string
return x
def default_capability_and_modified(self):
"""Set capability name and modified time in md
Every ResourceSync document should have these two top-level
metadata attributes.
"""
if ('capability' not in self.md and self.capability_md is not None):
self.md['capability']=self.capability_md
if ('modified' not in self.md):
self.md['modified']=datetime_to_str(no_fractions=True)
def head_on_file(self,file):
"""Mock up requests.head(..) response on local file
"""
response = HeadResponse()
if (not os.path.isfile(file)):
response.status_code='404'
else:
response.status_code='200'
response.headers['last-modified']=datetime_to_str(os.path.getmtime(file))
response.headers['content-length']=os.path.getsize(file)
return(response)
def incremental(self,
allow_deletion=False,
change_list_uri=None,
from_datetime=None):
"""Incremental synchronization
Use Change List to do incremental sync
"""
self.logger.debug("Starting incremental sync")
### 0. Sanity checks
if (len(self.mapper) < 1):
raise ClientFatalError(
"No source to destination mapping specified")
if (self.mapper.unsafe()):
raise ClientFatalError(
"Source to destination mappings unsafe: %s" % str(self.mapper))
from_timestamp = None
if (from_datetime is not None):
try:
from_timestamp = str_to_datetime(from_datetime)
except ValueError:
raise ClientFatalError("Bad datetime in --from (%s)" %
from_datetime)
### 1. Work out where to start from
if (from_timestamp is None):
from_timestamp = ClientState().get_state(self.sitemap)
if (from_timestamp is None):
raise ClientFatalError(
"Cannot do incremental sync. No stored timestamp for this site, and no explicit --from."
)
### 2. Get URI of change list, from sitemap or explicit
if (change_list_uri):
# Translate as necessary using maps
change_list = self.sitemap_uri(change_list_uri)
else:
# Try default name
change_list = self.sitemap_uri(self.change_list_name)
### 3. Read change list from source
try:
self.logger.info("Reading change list %s" % (change_list))
src_change_list = ChangeList()
src_change_list.read(uri=change_list)
self.logger.debug("Finished reading change list")
except Exception as e:
raise ClientFatalError(
"Can't read source change list from %s (%s)" %
(change_list, str(e)))
self.logger.info("Read source change list, %d changes listed" %
(len(src_change_list)))
#if (len(src_change_list)==0):
# raise ClientFatalError("Aborting as there are no resources to sync")
if (self.checksum and not src_change_list.has_md5()):
self.checksum = False
self.logger.info(
"Not calculating checksums on destination as not present in source change list"
)
# Check all changes have timestamp and record last
self.last_timestamp = 0
for resource in src_change_list:
if (resource.timestamp is None):
raise ClientFatalError(
"Aborting - missing timestamp for change in %s" % (uri))
if (resource.timestamp > self.last_timestamp):
self.last_timestamp = resource.timestamp
### 4. Check that the change list has authority over URIs listed
# FIXME - What does authority mean for change list? Here use both the
# change list URI and, if we used it, the sitemap URI
if (not self.noauth):
uauth_cs = UrlAuthority(change_list, self.strictauth)
if (not change_list_uri):
uauth_sm = UrlAuthority(self.sitemap)
for resource in src_change_list:
if (not uauth_cs.has_authority_over(resource.uri) and
(change_list_uri
or not uauth_sm.has_authority_over(resource.uri))):
raise ClientFatalError(
"Aborting as change list (%s) mentions resource at a location it does not have authority over (%s), override with --noauth"
% (change_list, resource.uri))
### 5. Prune entries before starting timestamp and dupe changes for a resource
num_skipped = src_change_list.prune_before(from_timestamp)
if (num_skipped > 0):
self.logger.info("Skipped %d changes before %s" %
(num_skipped, datetime_to_str(from_timestamp)))
num_dupes = src_change_list.prune_dupes()
if (num_dupes > 0):
self.logger.info("Removed %d prior changes" % (num_dupes))
# Review and log status before
# FIXME - should at this stage prune the change list to pick out
# only the last change for each resource
to_update = 0
to_create = 0
to_delete = 0
for resource in src_change_list:
if (resource.change == 'updated'):
to_update += 1
elif (resource.change == 'created'):
to_create += 1
elif (resource.change == 'deleted'):
to_delete += 1
else:
raise ClientError("Unknown change type %s" % (resource.change))
# Log status based on what we know from the Change List. Exit if
# either there are no changes or if there are only deletions and
# we don't allow deletion
in_sync = ((to_update + to_delete + to_create) == 0)
self.log_status(in_sync=in_sync,
incremental=True,
created=to_create,
updated=to_update,
deleted=to_delete)
if (in_sync or ((to_update + to_create) == 0 and not allow_deletion)):
self.logger.debug("Completed incremental")
return
### 6. Apply changes at same time or after from_timestamp
delete_msg = (", and delete %d resources" %
to_delete) if (allow_deletion) else ''
self.logger.warning("Will apply %d changes%s" %
(len(src_change_list), delete_msg))
num_updated = 0
num_deleted = 0
num_created = 0
for resource in src_change_list:
uri = resource.uri
file = self.mapper.src_to_dst(uri)
if (resource.change == 'updated'):
self.logger.info("updated: %s -> %s" % (uri, file))
self.update_resource(resource, file, 'updated')
num_updated += 1
elif (resource.change == 'created'):
self.logger.info("created: %s -> %s" % (uri, file))
self.update_resource(resource, file, 'created')
num_created += 1
elif (resource.change == 'deleted'):
num_deleted += self.delete_resource(resource, file,
allow_deletion)
else:
raise ClientError("Unknown change type %s" % (resource.change))
### 7. Report status and planned actions
self.log_status(incremental=True,
created=num_created,
updated=num_updated,
deleted=num_deleted,
to_delete=to_delete)
### 8. Record last timestamp we have seen
if (self.last_timestamp > 0):
ClientState().set_state(self.sitemap, self.last_timestamp)
self.logger.info("Written last timestamp %s for incremental sync" %
(datetime_to_str(self.last_timestamp)))
### 9. Done
self.logger.debug("Completed incremental sync")
def baseline_or_audit(self, allow_deletion=False, audit_only=False):
"""Baseline synchonization or audit
Both functions implemented in this routine because audit is a prerequisite
for a baseline sync. In the case of baseline sync the last timestamp seen
is recorded as client state.
"""
action = ('audit' if (audit_only) else 'baseline sync')
self.logger.debug("Starting " + action)
### 0. Sanity checks
if (len(self.mapper) < 1):
raise ClientFatalError(
"No source to destination mapping specified")
if (not audit_only and self.mapper.unsafe()):
raise ClientFatalError(
"Source to destination mappings unsafe: %s" % str(self.mapper))
### 1. Get inventories from both src and dst
# 1.a source resource list
try:
self.logger.info("Reading sitemap %s" % (self.sitemap))
src_resource_list = ResourceList(
allow_multifile=self.allow_multifile, mapper=self.mapper)
src_resource_list.read(uri=self.sitemap)
self.logger.debug("Finished reading sitemap")
except Exception as e:
raise ClientFatalError(
"Can't read source resource list from %s (%s)" %
(self.sitemap, str(e)))
self.logger.info("Read source resource list, %d resources listed" %
(len(src_resource_list)))
if (len(src_resource_list) == 0):
raise ClientFatalError(
"Aborting as there are no resources to sync")
if (self.checksum and not src_resource_list.has_md5()):
self.checksum = False
self.logger.info(
"Not calculating checksums on destination as not present in source resource list"
)
# 1.b destination resource list mapped back to source URIs
rlb = ResourceListBuilder(set_md5=self.checksum, mapper=self.mapper)
dst_resource_list = rlb.from_disk()
### 2. Compare these resource lists respecting any comparison options
(same, updated, deleted,
created) = dst_resource_list.compare(src_resource_list)
### 3. Report status and planned actions
self.log_status(in_sync=(len(updated) + len(deleted) +
len(created) == 0),
audit=True,
same=len(same),
created=len(created),
updated=len(updated),
deleted=len(deleted))
if (audit_only or len(created) + len(updated) + len(deleted) == 0):
self.logger.debug("Completed " + action)
return
### 4. Check that sitemap has authority over URIs listed
if (not self.noauth):
uauth = UrlAuthority(self.sitemap, strict=self.strictauth)
for resource in src_resource_list:
if (not uauth.has_authority_over(resource.uri)):
raise ClientFatalError(
"Aborting as sitemap (%s) mentions resource at a location it does not have authority over (%s), override with --noauth"
% (self.sitemap, resource.uri))
### 5. Grab files to do sync
delete_msg = (", and delete %d resources" %
len(deleted)) if (allow_deletion) else ''
self.logger.warning("Will GET %d resources%s" %
(len(created) + len(updated), delete_msg))
self.last_timestamp = 0
num_created = 0
num_updated = 0
num_deleted = 0
for resource in created:
uri = resource.uri
file = self.mapper.src_to_dst(uri)
self.logger.info("created: %s -> %s" % (uri, file))
num_created += self.update_resource(resource, file, 'created')
for resource in updated:
uri = resource.uri
file = self.mapper.src_to_dst(uri)
self.logger.info("updated: %s -> %s" % (uri, file))
num_updated += self.update_resource(resource, file, 'updated')
for resource in deleted:
uri = resource.uri
file = self.mapper.src_to_dst(uri)
num_deleted += self.delete_resource(resource, file, allow_deletion)
### 6. Store last timestamp to allow incremental sync
if (not audit_only and self.last_timestamp > 0):
ClientState().set_state(self.sitemap, self.last_timestamp)
self.logger.info("Written last timestamp %s for incremental sync" %
(datetime_to_str(self.last_timestamp)))
### 7. Done
self.log_status(in_sync=(len(updated) + len(deleted) +
len(created) == 0),
same=len(same),
created=num_created,
updated=num_updated,
deleted=num_deleted,
to_delete=len(deleted))
self.logger.debug("Completed %s" % (action))
def lastmod(self):
"""The Last-Modified data in W3C Datetime syntax, Z notation"""
if self.timestamp is None:
return None
return datetime_to_str(self.timestamp)
def incremental(self, allow_deletion=False, change_list_uri=None, from_datetime=None):
"""Incremental synchronization
"""
self.logger.debug("Starting incremental sync")
### 0. Sanity checks
if (len(self.mappings)<1):
raise ClientFatalError("No source to destination mapping specified")
from_timestamp = None
if (from_datetime is not None):
try:
from_timestamp = str_to_datetime(from_datetime)
except ValueError:
raise ClientFatalError("Bad datetime in --from (%s)" % from_datetime)
### 1. Work out where to start from
if (from_timestamp is None):
from_timestamp=ClientState().get_state(self.sitemap)
if (from_timestamp is None):
raise ClientFatalError("No stored timestamp for this site, and no explicit --from")
### 2. Get URI of change list, from sitemap or explicit
if (change_list_uri):
# Translate as necessary using maps
change_list = self.sitemap_uri(change_list_uri)
else:
# Try default name
change_list = self.sitemap_uri(self.change_list_name)
### 3. Read change list from source
try:
self.logger.info("Reading change list %s" % (change_list))
src_change_list = ChangeList()
src_change_list.read(uri=change_list)
self.logger.debug("Finished reading change list")
except Exception as e:
raise ClientFatalError("Can't read source change list from %s (%s)" % (change_list,str(e)))
self.logger.info("Read source change list, %d changes listed" % (len(src_change_list)))
#if (len(src_change_list)==0):
# raise ClientFatalError("Aborting as there are no resources to sync")
if (self.checksum and not src_change_list.has_md5()):
self.checksum=False
self.logger.info("Not calculating checksums on destination as not present in source change list")
# Check all changes have timestamp and record last
self.last_timestamp = 0
for resource in src_change_list:
if (resource.timestamp is None):
raise ClientFatalError("Aborting - missing timestamp for change in %s" % (uri))
if (resource.timestamp > self.last_timestamp):
self.last_timestamp = resource.timestamp
### 4. Check that the change list has authority over URIs listed
# FIXME - What does authority mean for change list? Here use both the
# change list URI and, if we used it, the sitemap URI
uauth_cs = UrlAuthority(change_list)
if (not change_list_uri):
uauth_sm = UrlAuthority(self.sitemap)
for resource in src_change_list:
if (not uauth_cs.has_authority_over(resource.uri) and
(change_list_uri or not uauth_sm.has_authority_over(resource.uri))):
if (self.noauth):
#self.logger.info("Change list (%s) mentions resource at a location it does not have authority over (%s)" % (change_list,resource.uri))
pass
else:
raise ClientFatalError("Aborting as change list (%s) mentions resource at a location it does not have authority over (%s), override with --noauth" % (change_list,resource.uri))
### 5. Prune entries before starting timestamp and dupe changes for a resource
num_skipped = src_change_list.prune_before(from_timestamp)
if (num_skipped>0):
self.logger.info("Skipped %d changes before %s" % (num_skipped,datetime_to_str(from_timestamp)))
num_dupes = src_change_list.prune_dupes()
if (num_dupes>0):
self.logger.info("Removed %d prior changes" % (num_dupes))
### 6. Apply changes at same time or after from_timestamp
self.logger.info("Applying %d changes" % (len(src_change_list)))
num_updated = 0
num_deleted = 0
num_created = 0
for resource in src_change_list:
uri = resource.uri
file = self.mapper.src_to_dst(uri)
if (resource.change == 'updated'):
self.logger.info("updated: %s -> %s" % (uri,file))
self.update_resource(resource,file,'updated')
num_updated+=1
elif (resource.change == 'created'):
self.logger.info("created: %s -> %s" % (uri,file))
self.update_resource(resource,file,'created')
num_created+=1
elif (resource.change == 'deleted'):
self.delete_resource(resource,file,allow_deletion)
num_deleted+=1
else:
raise ClientError("Unknown change type %s" % (resource.change) )
### 7. Report status and planned actions
self.log_status(in_sync=((num_updated+num_deleted+num_created)==0),
incremental=True,created=num_created, updated=num_updated,
deleted=num_deleted)
### 8. Record last timestamp we have seen
if (self.last_timestamp>0):
ClientState().set_state(self.sitemap,self.last_timestamp)
self.logger.info("Written last timestamp %s for incremental sync" % (datetime_to_str(self.last_timestamp)))
### 9. Done
self.logger.debug("Completed incremental sync")
def baseline_or_audit(self, allow_deletion=False, audit_only=False):
"""Baseline synchonization or audit
Both functions implemented in this routine because audit is a prerequisite
for a baseline sync. In the case of baseline sync the last timestamp seen
is recorded as client state.
"""
action = ( 'audit' if (audit_only) else 'baseline sync' )
self.logger.debug("Starting "+action)
### 0. Sanity checks
if (len(self.mappings)<1):
raise ClientFatalError("No source to destination mapping specified")
### 1. Get inventories from both src and dst
# 1.a source resource_list
try:
self.logger.info("Reading sitemap %s" % (self.sitemap))
src_resource_list = ResourceList(allow_multifile=self.allow_multifile, mapper=self.mapper)
src_resource_list.read(uri=self.sitemap)
self.logger.debug("Finished reading sitemap")
except Exception as e:
raise ClientFatalError("Can't read source resource_list from %s (%s)" % (self.sitemap,str(e)))
self.logger.info("Read source resource_list, %d resources listed" % (len(src_resource_list)))
if (len(src_resource_list)==0):
raise ClientFatalError("Aborting as there are no resources to sync")
if (self.checksum and not src_resource_list.has_md5()):
self.checksum=False
self.logger.info("Not calculating checksums on destination as not present in source resource_list")
# 1.b destination resource_list mapped back to source URIs
rlb = ResourceListBuilder(mapper=self.mapper)
rlb.do_md5=self.checksum
dst_resource_list = rlb.from_disk()
### 2. Compare these resource_lists respecting any comparison options
(same,updated,deleted,created)=dst_resource_list.compare(src_resource_list)
### 3. Report status and planned actions
self.log_status(in_sync=(len(updated)+len(deleted)+len(created)==0),
audit=True,same=len(same),created=len(created),
updated=len(updated),deleted=len(deleted))
if (audit_only or len(created)+len(updated)+len(deleted)==0):
self.logger.debug("Completed "+action)
return
### 4. Check that sitemap has authority over URIs listed
uauth = UrlAuthority(self.sitemap)
for resource in src_resource_list:
if (not uauth.has_authority_over(resource.uri)):
if (self.noauth):
#self.logger.info("Sitemap (%s) mentions resource at a location it does not have authority over (%s)" % (self.sitemap,resource.uri))
pass
else:
raise ClientFatalError("Aborting as sitemap (%s) mentions resource at a location it does not have authority over (%s), override with --noauth" % (self.sitemap,resource.uri))
### 5. Grab files to do sync
delete_msg = (", and delete %d resources" % len(deleted)) if (allow_deletion) else ''
self.logger.warning("Will GET %d resources%s" % (len(created)+len(updated),delete_msg))
self.last_timestamp = 0
num_created=0
num_updated=0
num_deleted=0
for resource in created:
uri = resource.uri
file = self.mapper.src_to_dst(uri)
self.logger.info("created: %s -> %s" % (uri,file))
num_created+=self.update_resource(resource,file,'created')
for resource in updated:
uri = resource.uri
file = self.mapper.src_to_dst(uri)
self.logger.info("updated: %s -> %s" % (uri,file))
num_updated+=self.update_resource(resource,file,'updated')
for resource in deleted:
uri = resource.uri
file = self.mapper.src_to_dst(uri)
num_deleted+=self.delete_resource(resource,file,allow_deletion)
### 6. Store last timestamp to allow incremental sync
if (not audit_only and self.last_timestamp>0):
ClientState().set_state(self.sitemap,self.last_timestamp)
self.logger.info("Written last timestamp %s for incremental sync" % (datetime_to_str(self.last_timestamp)))
### 7. Done
self.log_status(in_sync=(len(updated)+len(deleted)+len(created)==0),
same=len(same),created=num_created,
updated=num_updated,deleted=num_deleted)
self.logger.debug("Completed %s" % (action))
def md_until(self):
"""md_until value in W3C Datetime syntax, Z notation"""
return datetime_to_str(self._get_extra('ts_until'))
def md_from(self):
"""md_from value in W3C Datetime syntax, Z notation"""
return datetime_to_str(self._get_extra('ts_from'))
def md_completed(self):
"""md_completed value in W3C Datetime syntax, Z notation"""
return datetime_to_str(self._get_extra('ts_completed'))
def md_at(self):
"""md_at values in W3C Datetime syntax, Z notation"""
return datetime_to_str(self._get_extra('ts_at'))
def lastmod(self):
"""The Last-Modified data in W3C Datetime syntax, Z notation"""
return datetime_to_str(self.timestamp)