def ensure_folder(pathname):
tgt_path = tgt_folder
if not pathname.startswith(tgt_path):
return False
if hsds_global:
endpoint = hsds_global
else:
endpoint = None
folder = h5pyd.Folder(tgt_path,
endpoint=endpoint,
username=username,
password=password,
bucket=tgt_bucket)
names = pathname[len(tgt_path):].split("/")
for name in names:
if not name:
continue
tgt_path = os.path.join(tgt_path, name) + '/'
if name not in folder:
print(f"creating folder: {tgt_path}")
folder = h5pyd.Folder(tgt_path,
mode="w",
endpoint=endpoint,
username=username,
password=password,
bucket=tgt_bucket)
return True
def getHomeFolder(username):
if not username:
return None
dir = h5pyd.Folder('/home/') # get folder object for root
homefolder = None
for name in dir:
# we should come across the given domain
if username.startswith(name):
# check any folders where the name matches at least part of the username
# e.g. folder: "/home/bob/" for username "*****@*****.**"
path = '/home/' + name + '/'
try:
f = h5pyd.Folder(path)
except IOError as ioe:
print("got ioe:", ioe)
continue
except Exception as e:
print("got exception:", e)
continue
if f.owner == username:
homefolder = path
f.close()
if homefolder:
break
dir.close()
return homefolder
def check_res_file(res_file):
"""
Check resource to see if the given path
- It belongs to a multi-file handler
- Is on local disk
- Is a hsds path
Parameters
----------
res_file : str
Filepath to single resource file, unix style multi-file path like
/h5_dir/prefix*suffix.h5, or an hsds filepath (filename of hsds
path can also contain wildcards *)
Returns
-------
multi_h5_res : bool
Boolean flag to use a MultiFileResource handler
hsds : bool
Boolean flag to use h5pyd to handle .h5 'files' hosted on AWS
behind HSDS
"""
multi_h5_res = False
hsds = False
if os.path.isfile(res_file):
pass
elif '*' in res_file:
multi_h5_res = True
elif os.path.isdir(res_file):
msg = ('Cannot parse directory, need to add wildcard * suffix: {}'.
format(res_file))
raise FileInputError(msg)
else:
try:
import h5pyd
hsds_dir = os.path.dirname(res_file)
with h5pyd.Folder(hsds_dir + '/') as f:
hsds = True
fps = [
f'{hsds_dir}/{fn}' for fn in f
if fnmatch(f'{hsds_dir}/{fn}', res_file)
]
if not any(fps):
msg = (
'{} is not a valid HSDS file path!'.format(res_file))
raise FileNotFoundError(msg)
elif len(fps) > 1:
multi_h5_res = True
except Exception as ex:
msg = ("{} is not a valid file path, and HSDS "
"cannot be check for a file at this path:{}!".format(
res_file, ex))
raise FileNotFoundError(msg) from ex
return multi_h5_res, hsds
def getFolder(domain, mode='r'):
username = cfg["hs_username"]
password = cfg["hs_password"]
endpoint = cfg["hs_endpoint"]
bucket = cfg["hs_bucket"]
dir = h5py.Folder(domain, mode=mode, endpoint=endpoint, username=username, password=password, bucket=bucket)
return dir
def getFolder(domain):
username = cfg["hs_username"]
password = cfg["hs_password"]
endpoint = cfg["hs_endpoint"]
#print("getFolder", domain)
dir = h5py.Folder(domain, endpoint=endpoint, username=username, password=password)
return dir
def getDomainInfo(domain, cfg):
""" get info about the domain and print """
username = cfg["hs_username"]
password = cfg["hs_password"]
endpoint = cfg["hs_endpoint"]
if domain.endswith('/'):
is_folder = True
obj_class = "Folder"
else:
is_folder = False
obj_class = "Domain"
try:
if domain.endswith('/'):
f = h5pyd.Folder(domain,
mode='r',
endpoint=endpoint,
username=username,
password=password,
use_cache=True)
else:
f = h5pyd.File(domain,
mode='r',
endpoint=endpoint,
username=username,
password=password,
use_cache=True)
except IOError as oe:
if oe.errno in (404, 410): # Not Found
sys.exit("domain: {} not found".format(domain))
elif oe.errno == 401: # Unauthorized
sys.exit("Authorization failure")
elif oe.errno == 403: # Forbidden
sys.exit("Not allowed")
else:
sys.exit("Unexpected error: {}".format(oe))
timestamp = datetime.fromtimestamp(int(f.modified))
if is_folder:
print("folder: {}".format(domain))
print(" owner: {}".format(f.owner))
print(" last modified: {}".format(timestamp))
else:
# report HDF objects (groups, datasets, and named datatypes) vs. allocated chunks
num_objects = f.num_groups + f.num_datatypes + f.num_datasets
num_chunks = f.num_objects - num_objects
print("domain: {}".format(domain))
print(" owner: {}".format(f.owner))
print(" id: {}".format(f.id.id))
print(" last modified: {}".format(timestamp))
print(" total_size: {}".format(format_size(f.total_size)))
print(" allocated_bytes: {}".format(format_size(f.allocated_bytes)))
print(" num objects: {}".format(num_objects))
print(" num chunks: {}".format(num_chunks))
f.close()
def createFolder(domain):
username = cfg["hs_username"]
password = cfg["hs_password"]
endpoint = cfg["hs_endpoint"]
#print("getFolder", domain)
owner = None
if "hs_owner" in cfg:
owner=cfg["hs_owner"]
dir = h5py.Folder(domain, mode='x', endpoint=endpoint, username=username, password=password, owner=owner)
return dir
def getFolder(domain, mode="r"):
username = cfg["hs_username"]
password = cfg["hs_password"]
endpoint = cfg["hs_endpoint"]
dir = h5pyd.Folder(domain,
endpoint=endpoint,
username=username,
password=password,
mode=mode)
return dir
def getFolder(domain):
username = cfg["hs_username"]
password = cfg["hs_password"]
endpoint = cfg["hs_endpoint"]
bucket = cfg["hs_bucket"]
pattern = cfg["pattern"]
query = cfg["query"]
batch_size = 100 # use smaller batchsize for interactively listing of large collections
dir = h5py.Folder(domain, endpoint=endpoint, username=username,
password=password, bucket=bucket, pattern=pattern, query=query, batch_size=batch_size)
return dir
def rm_objects(self):
if not self.username or not self.password or not self.endpoint:
return
if not self.temp_dir:
return
folder = h5pyd.Folder(self.temp_dir,
mode='a',
endpoint=self.endpoint,
username=self.username,
password=self.password)
if len(folder) == 0:
time.sleep(10) # allow time for HSDS to sync to
for name in folder:
del folder[name]
def getHomeFolder():
username = cfg["hs_username"]
password = cfg["hs_password"]
endpoint = cfg["hs_endpoint"]
if not username:
return None
dir = h5pyd.Folder('/home/',
username=username,
password=password,
endpoint=endpoint) # get folder object for root
homefolder = None
for name in dir:
# we should come across the given domain
if username.startswith(name):
# check any folders where the name matches at least part of the username
# e.g. folder: "/home/bob/" for username "*****@*****.**"
path = '/home/' + name + '/'
try:
f = h5pyd.Folder(path,
username=username,
password=password,
endpoint=endpoint)
except IOError as ioe:
logging.info("find home folder - got ioe: {}".format(ioe))
continue
except Exception as e:
logging.warn("find home folder - got exception: {}".format(e))
continue
if f.owner == username:
homefolder = path
f.close()
if homefolder:
break
dir.close()
return homefolder
def _get_hsds_file_paths(h5_path, hsds_kwargs=None):
"""
Get a list of h5 filepaths matching the h5_path specification from HSDS
Parameters
----------
h5_path : str
Unix shell style pattern path with * wildcards to multi-file
resource file sets. Files must have the same coordinates
but can have different datasets or time indexes.
hsds_kwargs : dict, optional
Dictionary of optional kwargs for h5pyd, e.g., bucket, username,
password, by default None
Returns
-------
file_paths : list
List of filepaths for this handler to handle.
"""
import h5pyd
if hsds_kwargs is None:
hsds_kwargs = {}
hsds_dir = os.path.dirname(h5_path)
fn = os.path.basename(h5_path)
if '*' in hsds_dir:
msg = ('HSDS path specifications cannot handle wildcards in the '
'directory name! The directory must be explicit but the '
'filename can have wildcards. This HSDS h5_path input '
'cannot be used: {}'.format(h5_path))
raise FileNotFoundError(msg)
if not fn:
msg = ('h5_path must be a unix shell style pattern with '
'wildcard * in order to find files, but received '
'directory specification: {}'.format(h5_path))
raise FileInputError(msg)
with h5pyd.Folder(hsds_dir + '/', **hsds_kwargs) as f:
file_paths = [
f'{hsds_dir}/{fn}' for fn in f
if fnmatch(f'{hsds_dir}/{fn}', h5_path)
]
return file_paths
def check_res_file(res_file):
"""
Check resource to see if the given path
- It belongs to a multi-file handler
- Is on local disk
- Is a hsds path
Parameters
----------
res_file : str
Filepath to single resource file, multi-h5 directory,
or /h5_dir/prefix*suffix
Returns
-------
multi_h5_res : bool
Boolean flag to use a MultiFileResource handler
hsds : bool
Boolean flag to use h5pyd to handle .h5 'files' hosted on AWS
behind HSDS
"""
multi_h5_res = False
hsds = False
if os.path.isdir(res_file) or ('*' in res_file):
multi_h5_res = True
else:
if not os.path.isfile(res_file):
try:
import h5pyd
hsds_dir, hsds_file = os.path.split(res_file)
with h5pyd.Folder(hsds_dir + '/') as f:
hsds = True
if hsds_file not in f:
msg = ('{} is not a valid HSDS file path!'.format(
res_file))
print(msg)
raise FileNotFoundError(msg)
except Exception as ex:
msg = ("{} is not a valid file path, and HSDS "
"cannot be check for a file at this path:{}!".format(
res_file, ex))
raise FileNotFoundError(msg)
return multi_h5_res, hsds
def __init__(self, loc, **kwargs):
"""A set of FIREfly domains that satisfy filter criteria.
Parameters
----------
loc : str
A Kita server URI where FIREfly flight data are hosted.
Other parameters
----------------
mode : str
Access mode. Default is "r".
pattern : str
A Python regex for filtering FIREfly flight file names.
query : str
A boolean expression for filtering FIREfly flights' data.
kwargs : dict
Any remaining named arguments are assumed to be flight data
filtering parameters or Kita server access information.
"""
self._mode = kwargs.pop('mode', 'r')
pattern = kwargs.pop('pattern', None)
query = kwargs.pop('query', None)
if query is None:
self._flight_filter, self._data_filter = filter_builder(kwargs)
else:
self._flight_filter = query
self._data_filter = None
self._loc = h5pyd.Folder(loc,
mode=self._mode,
pattern=pattern,
query=self._flight_filter,
**kwargs)
loc = self._loc.domain
self._domains = [loc + d for d in self._loc]
self._kwargs = kwargs
def main():
if len(sys.argv) == 1 or (sys.argv[1] == "-h" or sys.argv[1] == "--help"):
printUsage()
sys.exit(1)
bucket = None
if len(sys.argv) == 4 and sys.argv[1] == "--bucket":
bucket = sys.argv[2]
print("using bucket:", bucket)
folder_path = sys.argv[-1]
if folder_path[0] != '/' or folder_path[-1] != '/':
print("domain folder must start and end with '/")
printUsage()
folder = h5pyd.Folder(folder_path, mode='r+')
domains = []
for domain in folder:
domains.append(domain)
for domain in domains:
print(f"removing: {domain}")
del folder[domain]
# Utilities. The full HDF5 REST Server copyright notice, including # # terms governing use, modification, and redistribution, is contained in # # the file COPYING, which can be found at the root of the source code # # distribution tree. If you do not have access to this file, you may # # request a copy from [email protected]. # ############################################################################## import sys import os import config import h5pyd if config.get("use_h5py"): sys.exit("use_h5py") if "H5PYD_TEST_FOLDER" not in os.environ: sys.exit("set H5PYD_TEST_FOLDER environment not set") folder_path = os.environ["H5PYD_TEST_FOLDER"] if not folder_path.startswith("/"): # HSDS expects folder paths to start with a slash (as opposed to DNS format) sys.exit("not HSDS path") try: h5pyd.Folder(folder_path) # will trigger error with h5serv except Exception: sys.exit("Server doesn't support Folder objects")
if 'BasicFusionFiles' not in sdb.list_domains()['DomainNames']:
sdb.create_domain(DomainName="BasicFusionFiles")
if args.list:
q = sdb.select(
SelectExpression='select * from BasicFusionFiles where MISR_Path="P125"'
)
if 'Items' in q:
for item in q['Items']:
print(item)
sys.exit(0)
misr_paths = h5py.Folder(domain_name=args.domain,
endpoint=args.hsds_endpoint,
username=args.user,
password=args.password)
for path in misr_paths:
h5_files = h5py.Folder(domain_name=args.domain + path + "/",
endpoint=args.hsds_endpoint,
username=args.user,
password=args.password)
for h5_file in h5_files:
root_name = h5_file.split('.')[0]
satellite, dataset, level, orbit, timestamp, _, _ = root_name.split(
"_")
year = timestamp[:4]
month = timestamp[4:6]
day = timestamp[6:8]
print(year, month, day)
import sys
import subprocess
import h5pyd
HSDS_BUCKET = "firefly-hsds"
H5_FOLDER = "/FIREfly/h5/"
folder = h5pyd.Folder(H5_FOLDER, bucket=HSDS_BUCKET) # get folder object
domain_names = []
for domain in folder:
domain_names.append(domain)
if not domain_names:
print("no domains found!")
sys.exit()
print(f"found {len(domain_names)} domains")
for domain_name in domain_names:
domain = H5_FOLDER + domain_name
print(f"removing {domain}")
rc = subprocess.run(["hsrm", "--bucket", HSDS_BUCKET, domain])
if rc.returncode > 0:
print(f"unable to delete {domain}")
sys.exit(-1)
print("done!")
def getDomainInfo(domain, cfg):
""" get info about the domain and print """
username = cfg["hs_username"]
password = cfg["hs_password"]
endpoint = cfg["hs_endpoint"]
bucket = cfg["hs_bucket"]
if "rescan" in cfg and cfg["rescan"]:
mode = "r+" # need write intent
else:
mode = 'r'
if domain.endswith('/'):
is_folder = True
else:
is_folder = False
try:
if is_folder:
f = h5pyd.Folder(domain,
mode=mode,
endpoint=endpoint,
username=username,
password=password,
bucket=bucket,
use_cache=True)
else:
f = h5pyd.File(domain,
mode=mode,
endpoint=endpoint,
username=username,
password=password,
bucket=bucket,
use_cache=False)
except IOError as oe:
if oe.errno in (404, 410): # Not Found
sys.exit("domain: {} not found".format(domain))
elif oe.errno == 401: # Unauthorized
sys.exit("Authorization failure")
elif oe.errno == 403: # Forbidden
sys.exit("Not allowed")
else:
sys.exit("Unexpected error: {}".format(oe))
timestamp = datetime.fromtimestamp(int(f.modified))
if not is_folder and f.last_scan:
last_scan = datetime.fromtimestamp(int(f.last_scan))
else:
last_scan = None
if is_folder:
print("folder: {}".format(domain))
print(" owner: {}".format(f.owner))
print(" last modified: {}".format(timestamp))
else:
if "rescan" in cfg and cfg["rescan"]:
f.run_scan()
# report HDF objects (groups, datasets, and named datatypes) vs. allocated chunks
num_objects = f.num_groups + f.num_datatypes + f.num_datasets
if f.num_chunks > 0:
num_chunks = f.num_chunks
else:
# older storeinfo format doesn't have num_chunks, so calculate
num_chunks = f.num_objects - num_objects
print("domain: {}".format(domain))
print(" owner: {}".format(f.owner))
print(" id: {}".format(f.id.id))
print(" last modified: {}".format(timestamp))
if last_scan:
print(" last scan: {}".format(last_scan))
if f.md5_sum:
print(" md5 sum: {}".format(f.md5_sum))
print(" total_size: {}".format(format_size(f.total_size)))
print(" allocated_bytes: {}".format(format_size(f.allocated_bytes)))
if f.metadata_bytes:
print(" metadata_bytes: {}".format(
format_size(f.metadata_bytes)))
if f.linked_bytes:
print(" linked_bytes: {}".format(format_size(
f.linked_bytes)))
print(" num objects: {}".format(num_objects))
print(" num chunks: {}".format(num_chunks))
if f.num_linked_chunks:
print(" linked chunks: {}".format(f.num_linked_chunks))
f.close()
def main():
cfg["cmd"] = sys.argv[0].split('/')[-1]
if cfg["cmd"].endswith(".py"):
cfg["cmd"] = "python " + cfg["cmd"]
cfg["verbose"] = False
perm_abvr = {'c':'create', 'r': 'read', 'u': 'update', 'd': 'delete', 'e': 'readACL', 'p':'updateACL'}
fields = ('username', 'create', 'read', 'update', 'delete', 'readACL', 'updateACL')
domain = None
perm = None
loglevel = logging.ERROR
logfname = None
usernames = []
add_list = set()
remove_list = set()
if len(sys.argv) == 1 or sys.argv[1] == "-h":
printUsage()
# setup logging
logging.basicConfig(filename=logfname, format='%(asctime)s %(message)s', level=loglevel)
logging.debug("set log_level to {}".format(loglevel))
argn = 1
while argn < len(sys.argv):
arg = sys.argv[argn]
val = None
if len(sys.argv) > argn + 1:
val = sys.argv[argn+1]
logging.debug("arg:", arg, "val:", val)
if domain is None and arg in ("-v", "--verbose"):
cfg["verbose"] = True
argn += 1
elif domain is None and arg == "--loglevel":
val = val.upper()
if val == "DEBUG":
loglevel = logging.DEBUG
elif val == "INFO":
loglevel = logging.INFO
elif val in ("WARN", "WARNING"):
loglevel = logging.WARNING
elif val == "ERROR":
loglevel = logging.ERROR
else:
printUsage()
argn += 2
elif domain is None and arg == '--logfile':
logfname = val
argn += 2
elif domain is None and arg in ("-h", "--help"):
printUsage()
elif domain is None and arg in ("-e", "--endpoint"):
cfg["hs_endpoint"] = val
argn += 2
elif domain is None and arg in ("-u", "--username"):
cfg["hs_username"] = val
argn += 2
elif domain is None and arg in ("-p", "--password"):
cfg["hs_password"] = val
argn += 2
elif domain is None and arg[0] in ('-', '+'):
print("No domain given")
printUsage()
elif domain is None:
logging.debug("get domain")
domain = arg
if domain[0] != '/':
print("Domain must start with '/'")
printUsage()
logging.debug("domain:", domain)
argn += 1
elif arg[0] == '+':
logging.debug("got plus")
if len(usernames) > 0:
logging.debug("usernames:", usernames)
printUsage()
add_list = set(arg[1:])
logging.info("add_list:", add_list)
argn += 1
elif arg[0] == '-':
logging.debug("got minus")
if len(usernames) > 0:
printUsage()
remove_list = set(arg[1:])
logging.info("remove_list:", remove_list)
argn += 1
else:
logging.info("got username:"******"Invalid username:"******"domain:", domain)
logging.info("add_list:", add_list)
logging.info("remove_list:", remove_list)
logging.info("usernames:", usernames)
if len(usernames) == 0 and (add_list or remove_list):
print("At least one username must be given to add/remove permissions")
printUsage()
if domain is None:
print("no domain specified")
sys.exit(1)
conflicts = list(add_list & remove_list)
if len(conflicts) > 0:
print("permission: ", conflicts[0], " permission flag set for both add and remove")
sys.exit(1)
mode = 'r'
if add_list or remove_list:
mode = 'a' # we'll be updating the domain
perm = {}
for x in add_list:
if x not in perm_abvr:
print("Permission flag: {} is not valid - must be one of 'crudep;".format(x))
sys.exit(1)
perm_name = perm_abvr[x]
perm[perm_name] = True
for x in remove_list:
if x not in perm_abvr:
print("Permission flag: {} is not valid - must be one of 'crudep;".format(x))
sys.exit(1)
perm_name = perm_abvr[x]
perm[perm_name] = False
logging.info("perm:", perm)
# open the domain or folder
try:
if domain[-1] == '/':
f = h5pyd.Folder(domain, mode=mode, endpoint=cfg["hs_endpoint"], username=cfg["hs_username"], password=cfg["hs_password"])
else:
f = h5pyd.File(domain, mode=mode, endpoint=cfg["hs_endpoint"], username=cfg["hs_username"], password=cfg["hs_password"])
except IOError as ioe:
if ioe.errno in (404, 410):
print("domain not found")
sys.exit(1)
elif ioe.errno in (401, 403):
print("access is not authorized")
sys.exit(1)
else:
print("Unexpected error:", ioe)
sys.exit(1)
# update/add ACL if permission flags have been set
if perm:
default_acl = {'updateACL': False,
'delete': False,
'create': False,
'read': False,
'update': False,
'readACL': False,
'userName': '******'
}
# note: list.copy not supported in py2.7, copy by hand for now
# update_names = usernames.copy()
update_names = []
for username in usernames:
update_names.append(username)
if not update_names:
update_names.append("default")
for username in update_names:
# get user's ACL if it exist
acl = getACL(f, username=username)
if acl is None:
acl = default_acl.copy()
acl["userName"] = username
logging.info("updating acl to: {}".format(acl))
# mix in any permission changes
for k in perm:
acl[k] = perm[k]
try:
f.putACL(acl)
except IOError as ioe:
if ioe.errno in (401, 403):
print("access is not authorized")
else:
print("Unexpected error:", ioe)
sys.exit(1)
#
# read the acls
#
if len(usernames) == 0:
# no usernames, dump all ACLs
try:
acls = f.getACLs()
except IOError as ioe:
if ioe.errno == 403:
print("User {} does not have permission to read ACL for this domain".format(cfg["hs_username"]))
sys.exit(1)
elif ioe.errno == 401:
print("username/password needs to be provided")
sys.exit(1)
else:
print("Unexpected error: {}".format(ioe))
print("%015s %08s %08s %08s %08s %08s %08s " % fields)
print("-"*80)
for acl in acls:
vals = (acl["userName"], acl["create"], acl["read"], acl["update"], acl["delete"], acl["readACL"], acl["updateACL"])
print("%015s %08s %08s %08s %08s %08s %08s " % vals)
else:
header_printed = False # don't print header until we have at least one ACL
for username in usernames:
try:
acl = f.getACL(username)
if not header_printed:
print("%015s %08s %08s %08s %08s %08s %08s " % fields)
print("-"*80)
header_printed = True
vals = (acl["userName"], acl["create"], acl["read"], acl["update"], acl["delete"], acl["readACL"], acl["updateACL"])
print("%015s %08s %08s %08s %08s %08s %08s " % vals)
except IOError as ioe:
if ioe.errno == 403:
print("User {} does not have permission to read ACL for this domain".format(cfg["hs_username"]))
sys.exit(1)
elif ioe.errno == 401:
print("username/password needs to be provided")
sys.exit(1)
elif ioe.errno == 404:
print(username, "<NONE>")
else:
print("Unexpected error:", ioe)
sys.exit(1)
f.close()
# Main
#
print("load_file.py")
if "H5PYD_TEST_FOLDER" not in os.environ:
print("set H5PYD_TEST_FOLDER environment not set")
sys.exit(1)
test_folder = os.environ["H5PYD_TEST_FOLDER"]
data_dir = "data"
s3_http_path = "https://s3.amazonaws.com/hdfgroup/data/hdf5test/"
parent = h5pyd.Folder(test_folder)
filenames = config.get_test_filenames()
if not os.path.exists(data_dir):
# make data directory for downloaded HDF5 files
os.mkdir(data_dir)
for filename in filenames:
print(filename)
domain_path = os.path.join(test_folder, filename)
print(domain_path)
if filename in parent:
print("found")
continue
# check to see if the file has already been downloaded
hdf5_path = os.path.join(data_dir, filename)
