async def getChunk(app, chunk_id, dset_json, s3path=None, s3offset=0, s3size=0, chunk_init=False):
# if the chunk cache has too many dirty items, wait till items get flushed to S3
MAX_WAIT_TIME = 10.0 # TBD - make this a config
chunk_cache = app['chunk_cache']
if chunk_init and s3offset > 0:
log.error(f"unable to initiale chunk {chunk_id} for reference layouts ")
raise HTTPInternalServerError()
log.debug(f"getChunk cache utilization: {chunk_cache.cacheUtilizationPercent} per, dirty_count: {chunk_cache.dirtyCount}, mem_dirty: {chunk_cache.memDirty}")
chunk_arr = None
dset_id = getDatasetId(chunk_id)
dims = getChunkLayout(dset_json)
type_json = dset_json["type"]
dt = createDataType(type_json)
bucket = None
s3key = None
if s3path:
if not s3path.startswith("s3://"):
# TBD - verify these at dataset creation time?
log.error(f"unexpected s3path for getChunk: {s3path}")
raise HTTPInternalServerError()
path = s3path[5:]
index = path.find('/') # split bucket and key
if index < 1:
log.error(f"s3path is invalid: {s3path}")
raise HTTPInternalServerError()
bucket = path[:index]
log.debug(f"using bucket: {bucket}")
s3key = path[(index+1):]
log.debug(f"Using bucket: {bucket} and s3key: {s3key}")
else:
s3key = getS3Key(chunk_id)
log.debug("getChunk s3key: {}".format(s3key))
if chunk_id in chunk_cache:
chunk_arr = chunk_cache[chunk_id]
else:
if s3path and s3size == 0:
obj_exists = False
else:
obj_exists = await isS3Obj(app, s3key, bucket=bucket)
# TBD - potential race condition?
if obj_exists:
pending_s3_read = app["pending_s3_read"]
if chunk_id in pending_s3_read:
# already a read in progress, wait for it to complete
read_start_time = pending_s3_read[chunk_id]
log.info(f"s3 read request for {chunk_id} was requested at: {read_start_time}")
while time.time() - read_start_time < 2.0:
log.debug("waiting for pending s3 read, sleeping")
await asyncio.sleep(1) # sleep for sub-second?
if chunk_id in chunk_cache:
log.info(f"Chunk {chunk_id} has arrived!")
chunk_arr = chunk_cache[chunk_id]
break
if chunk_arr is None:
log.warn(f"s3 read for chunk {chunk_id} timed-out, initiaiting a new read")
if chunk_arr is None:
if chunk_id not in pending_s3_read:
pending_s3_read[chunk_id] = time.time()
log.debug("Reading chunk {} from S3".format(s3key))
deflate_level = getDeflate(app, dset_id, dset_json)
chunk_bytes = await getS3Bytes(app, s3key, deflate_level=deflate_level, s3offset=s3offset, s3size=s3size, bucket=bucket)
if chunk_id in pending_s3_read:
# read complete - remove from pending map
elapsed_time = time.time() - pending_s3_read[chunk_id]
log.info(f"s3 read for {s3key} took {elapsed_time}")
del pending_s3_read[chunk_id]
else:
log.warn(f"expected to find {chunk_id} in pending_s3_read map")
chunk_arr = np.fromstring(chunk_bytes, dtype=dt)
chunk_arr = chunk_arr.reshape(dims)
log.debug("chunk size: {}".format(chunk_arr.size))
elif chunk_init:
log.debug("Initializing chunk {chunk_id}")
fill_value = getFillValue(dset_json)
if fill_value:
# need to convert list to tuples for numpy broadcast
if isinstance(fill_value, list):
fill_value = tuple(fill_value)
chunk_arr = np.empty(dims, dtype=dt, order='C')
chunk_arr[...] = fill_value
else:
chunk_arr = np.zeros(dims, dtype=dt, order='C')
else:
log.debug(f"Chunk {chunk_id} not found")
if chunk_arr is not None:
# check that there's room in the cache before adding it
if chunk_cache.memTarget - chunk_cache.memDirty < chunk_arr.size:
# no room in the cache, wait till space is freed by the s3sync task
wait_start = time.time()
while chunk_cache.memTarget - chunk_cache.memDirty < chunk_arr.size:
log.warn(f"getChunk, cache utilization: {chunk_cache.cacheUtilizationPercent}, sleeping till items are flushed")
if time.time() - wait_start > MAX_WAIT_TIME:
log.error(f"unable to save updated chunk {chunk_id} to cache returning 503 error")
raise HTTPServiceUnavailable()
await asyncio.sleep(1)
chunk_cache[chunk_id] = chunk_arr # store in cache
return chunk_arr
async def GET_Chunk(request):
log.request(request)
app = request.app
params = request.rel_url.query
chunk_id = request.match_info.get('id')
if not chunk_id:
msg = "Missing chunk id"
log.error(msg)
raise HTTPBadRequest(reason=msg)
if not isValidUuid(chunk_id, "Chunk"):
msg = "Invalid chunk id: {}".format(chunk_id)
log.warn(msg)
raise HTTPBadRequest(reason=msg)
validateInPartition(app, chunk_id)
log.debug("request params: {}".format(list(params.keys())))
if "dset" not in params:
msg = "Missing dset in GET request"
log.error(msg)
raise HTTPBadRequest(reason=msg)
dset_json = json.loads(params["dset"])
log.debug("dset_json: {}".format(dset_json))
type_json = dset_json["type"]
dims = getChunkLayout(dset_json)
log.debug("got dims: {}".format(dims))
rank = len(dims)
# get chunk selection from query params
if "select" in params:
log.debug("select: {}".format(params["select"]))
selection = []
for i in range(rank):
dim_slice = getSliceQueryParam(request, i, dims[i])
selection.append(dim_slice)
selection = tuple(selection)
log.debug("got selection: {}".format(selection))
dt = createDataType(type_json)
log.debug("dtype: {}".format(dt))
rank = len(dims)
if rank == 0:
msg = "No dimension passed to GET chunk request"
log.error(msg)
raise HTTPBadRequest(reason=msg)
if len(selection) != rank:
msg = "Selection rank does not match shape rank"
log.error(msg)
raise HTTPBadRequest(reason=msg)
for i in range(rank):
s = selection[i]
log.debug("selection[{}]: {}".format(i, s))
s3path = None
s3offset = 0
s3size = 0
if "s3path" in params:
s3path = params["s3path"]
log.debug(f"GET_Chunk - useing s3path: {s3path}")
if "s3offset" in params:
try:
s3offset = int(params["s3offset"])
except ValueError:
log.error(f"invalid s3offset params: {params['s3offset']}")
raise HTTPBadRequest()
if "s3size" in params:
try:
s3size = int(params["s3size"])
except ValueError:
log.error(f"invalid s3size params: {params['s3sieze']}")
raise HTTPBadRequest()
chunk_arr = await getChunk(app, chunk_id, dset_json, s3path=s3path, s3offset=s3offset, s3size=s3size)
if chunk_arr is None:
# return a 404
msg = "Chunk {} does not exist".format(chunk_id)
log.info(msg)
raise HTTPNotFound()
resp = None
if "query" in params:
# do query selection
query = params["query"]
log.info("query: {}".format(query))
if rank != 1:
msg = "Query selection only supported for one dimensional arrays"
log.warn(msg)
raise HTTPBadRequest(reason=msg)
limit = 0
if "Limit" in params:
limit = int(params["Limit"])
values = []
indices = []
field_names = []
if dt.fields:
field_names = list(dt.fields.keys())
x = chunk_arr[selection]
log.debug("x: {}".format(x))
eval_str = getEvalStr(query, "x", field_names)
log.debug("eval_str: {}".format(eval_str))
where_result = np.where(eval(eval_str))
log.debug("where_result: {}".format(where_result))
where_result_index = where_result[0]
log.debug("whare_result index: {}".format(where_result_index))
log.debug("boolean selection: {}".format(x[where_result_index]))
s = selection[0]
count = 0
for index in where_result_index:
log.debug("index: {}".format(index))
value = x[index].tolist()
log.debug("value: {}".format(value))
json_val = bytesArrayToList(value)
log.debug("json_value: {}".format(json_val))
json_index = index.tolist() * s.step + s.start # adjust for selection
indices.append(json_index)
values.append(json_val)
count += 1
if limit > 0 and count >= limit:
log.info("got limit items")
break
query_result = {}
query_result["index"] = indices
query_result["value"] = values
log.info(f"query_result retiurning: {len(indices)} rows")
log.debug(f"query_result: {query_result}")
resp = json_response(query_result)
else:
# get requested data
output_arr = chunk_arr[selection]
output_data = arrayToBytes(output_arr)
# write response
try:
resp = StreamResponse()
resp.headers['Content-Type'] = "application/octet-stream"
resp.content_length = len(output_data)
await resp.prepare(request)
await resp.write(output_data)
except Exception as e:
log.error(f"Exception during binary data write: {e}")
raise HTTPInternalServerError()
finally:
await resp.write_eof()
return resp
async def POST_Chunk(request):
log.request(request)
app = request.app
params = request.rel_url.query
put_points = False
num_points = 0
if "count" in params:
num_points = int(params["count"])
if "action" in params and params["action"] == "put":
log.info("POST Chunk put points, num_points: {}".format(num_points))
put_points = True
else:
log.info("POST Chunk get points")
chunk_id = request.match_info.get('id')
if not chunk_id:
msg = "Missing chunk id"
log.error(msg)
raise HTTPBadRequest(reason=msg)
log.info("POST chunk_id: {}".format(chunk_id))
chunk_index = getChunkIndex(chunk_id)
log.debug("chunk_index: {}".format(chunk_index))
if not isValidUuid(chunk_id, "Chunk"):
msg = "Invalid chunk id: {}".format(chunk_id)
log.warn(msg)
raise HTTPBadRequest(reason=msg)
validateInPartition(app, chunk_id)
log.debug("request params: {}".format(list(params.keys())))
if "dset" not in params:
msg = "Missing dset in GET request"
log.error(msg)
raise HTTPBadRequest(reason=msg)
dset_json = json.loads(params["dset"])
log.debug("dset_json: {}".format(dset_json))
chunk_layout = getChunkLayout(dset_json)
chunk_coord = getChunkCoordinate(chunk_id, chunk_layout)
log.debug("chunk_coord: {}".format(chunk_coord))
if not request.has_body:
msg = "POST Value with no body"
log.warn(msg)
raise HTTPBadRequest(reason=msg)
content_type = "application/octet-stream"
if "Content-Type" in request.headers:
# client should use "application/octet-stream" for binary transfer
content_type = request.headers["Content-Type"]
if content_type != "application/octet-stream":
msg = "Unexpected content_type: {}".format(content_type)
log.error(msg)
raise HTTPBadRequest(reason=msg)
type_json = dset_json["type"]
dset_dtype = createDataType(type_json)
log.debug("dtype: {}".format(dset_dtype))
dims = getChunkLayout(dset_json)
log.debug("got dims: {}".format(dims))
rank = len(dims)
if rank == 0:
msg = "POST chunk request with no dimensions"
log.error(msg)
raise HTTPBadRequest(reason=msg)
# create a numpy array for incoming points
input_bytes = await request_read(request)
if len(input_bytes) != request.content_length:
msg = "Read {} bytes, expecting: {}".format(len(input_bytes), request.content_length)
log.error(msg)
raise HTTPInternalServerError()
# get chunk from cache/s3. If not found init a new chunk if this is a write request
chunk_arr = await getChunk(app, chunk_id, dset_json, chunk_init=put_points)
if put_points:
# writing point data
# create a numpy array with the following type:
# (coord1, coord2, ...) | dset_dtype
if rank == 1:
coord_type_str = "uint64"
else:
coord_type_str = "({},)uint64".format(rank)
comp_dtype = np.dtype([("coord", np.dtype(coord_type_str)), ("value", dset_dtype)])
point_arr = np.fromstring(input_bytes, dtype=comp_dtype)
if len(point_arr) != num_points:
msg = "Unexpected size of point array, got: {} expected: {}".format(len(point_arr), num_points)
log.warn(msg)
raise HTTPBadRequest(reason=msg)
for i in range(num_points):
elem = point_arr[i]
if rank == 1:
coord = int(elem[0])
else:
coord = tuple(elem[0]) # index to update
val = elem[1] # value
chunk_arr[coord] = val # update the point
chunk_cache = app["chunk_cache"]
chunk_cache.setDirty(chunk_id)
# async write to S3
dirty_ids = app["dirty_ids"]
now = int(time.time())
dirty_ids[chunk_id] = now
log.info("set {} to dirty".format(chunk_id))
else:
# reading point data
point_dt = np.dtype('uint64') # use unsigned long for point index
point_arr = np.fromstring(input_bytes, dtype=point_dt) # read points as unsigned longs
if len(point_arr) % rank != 0:
msg = "Unexpected size of point array"
log.warn(msg)
raise HTTPBadRequest(reason=msg)
num_points = len(point_arr) // rank
log.debug("got {} points".format(num_points))
point_arr = point_arr.reshape((num_points, rank))
output_arr = np.zeros((num_points,), dtype=dset_dtype)
for i in range(num_points):
point = point_arr[i,:]
tr_point = getChunkRelativePoint(chunk_coord, point)
val = chunk_arr[tuple(tr_point)]
output_arr[i] = val
if put_points:
# write empty response
resp = json_response({})
else:
# get data
output_data = output_arr.tobytes()
# write response
try:
resp = StreamResponse()
resp.headers['Content-Type'] = "application/octet-stream"
resp.content_length = len(output_data)
await resp.prepare(request)
await resp.write(output_data)
except Exception as e:
log.error(f"Exception during binary data write: {e}")
raise HTTPInternalServerError()
finally:
await resp.write_eof()
return resp
async def POST_Chunk(request):
log.request(request)
app = request.app
params = request.rel_url.query
put_points = False
num_points = 0
if "count" in params:
num_points = int(params["count"])
if "action" in params and params["action"] == "put":
log.info(f"POST Chunk put points, num_points: {num_points}")
put_points = True
else:
log.info("POST Chunk get points")
s3path = None
s3offset = 0
s3size = 0
if "s3path" in params:
if put_points:
log.error("s3path can not be used with put points POST request")
raise HTTPBadRequest()
s3path = params["s3path"]
log.debug(f"GET_Chunk - using s3path: {s3path}")
bucket = None
elif "bucket" in params:
bucket = params["bucket"]
else:
bucket = None
if "s3offset" in params:
try:
s3offset = int(params["s3offset"])
except ValueError:
log.error(f"invalid s3offset params: {params['s3offset']}")
raise HTTPBadRequest()
if "s3size" in params:
try:
s3size = int(params["s3size"])
except ValueError:
log.error(f"invalid s3size params: {params['s3sieze']}")
raise HTTPBadRequest()
chunk_id = request.match_info.get('id')
if not chunk_id:
msg = "Missing chunk id"
log.error(msg)
raise HTTPBadRequest(reason=msg)
log.info(f"POST chunk_id: {chunk_id}")
chunk_index = getChunkIndex(chunk_id)
log.debug(f"chunk_index: {chunk_index}")
if not isValidUuid(chunk_id, "Chunk"):
msg = f"Invalid chunk id: {chunk_id}"
log.warn(msg)
raise HTTPBadRequest(reason=msg)
validateInPartition(app, chunk_id)
log.debug(f"request params: {list(params.keys())}")
if "dset" in params:
msg = "Unexpected dset in POST request"
log.error(msg)
raise HTTPBadRequest(reason=msg)
dset_id = getDatasetId(chunk_id)
dset_json = await get_metadata_obj(app, dset_id, bucket=bucket)
log.debug(f"dset_json: {dset_json}")
chunk_layout = getChunkLayout(dset_json)
chunk_coord = getChunkCoordinate(chunk_id, chunk_layout)
log.debug(f"chunk_coord: {chunk_coord}")
if not request.has_body:
msg = "POST Value with no body"
log.warn(msg)
raise HTTPBadRequest(reason=msg)
content_type = "application/octet-stream"
if "Content-Type" in request.headers:
# client should use "application/octet-stream" for binary transfer
content_type = request.headers["Content-Type"]
if content_type != "application/octet-stream":
msg = f"Unexpected content_type: {content_type}"
log.error(msg)
raise HTTPBadRequest(reason=msg)
type_json = dset_json["type"]
dset_dtype = createDataType(type_json)
log.debug(f"dtype: {dset_dtype}")
dims = getChunkLayout(dset_json)
log.debug(f"got dims: {dims}")
rank = len(dims)
if rank == 0:
msg = "POST chunk request with no dimensions"
log.error(msg)
raise HTTPBadRequest(reason=msg)
# create a numpy array for incoming points
input_bytes = await request_read(request)
if len(input_bytes) != request.content_length:
msg = f"Read {len(input_bytes)} bytes, expecting: {request.content_length}"
log.error(msg)
raise HTTPInternalServerError()
# get chunk from cache/s3. If not found init a new chunk if this is a write request
chunk_arr = await getChunk(app,
chunk_id,
dset_json,
bucket=bucket,
s3path=s3path,
s3offset=s3offset,
s3size=s3size,
chunk_init=put_points)
if chunk_arr is None:
if put_points:
log.error("no array returned for put_points")
raise HTTPInternalServerError()
else:
# get points on a non-existent S3 objects?
log.warn("S3 object not found for get points")
raise HTTPNotFound()
log.debug(f"chunk_arr.shape: {chunk_arr.shape}")
if put_points:
# writing point data
# create a numpy array with the following type:
# (coord1, coord2, ...) | dset_dtype
if rank == 1:
coord_type_str = "uint64"
else:
coord_type_str = f"({rank},)uint64"
comp_dtype = np.dtype([("coord", np.dtype(coord_type_str)),
("value", dset_dtype)])
point_arr = np.fromstring(input_bytes, dtype=comp_dtype)
if len(point_arr) != num_points:
msg = f"Unexpected size of point array, got: {len(point_arr)} expected: {num_points}"
log.warn(msg)
raise HTTPBadRequest(reason=msg)
for i in range(num_points):
elem = point_arr[i]
log.debug(f"non-relative coordinate: {elem}")
if rank == 1:
coord = int(elem[0])
coord = coord % chunk_layout[0] # adjust to chunk relative
else:
coord = elem[0] # index to update
for dim in range(rank):
# adjust to chunk relative
coord[dim] = int(coord[dim]) % chunk_layout[dim]
coord = tuple(coord) # need to convert to a tuple
log.debug(f"relative coordinate: {coord}")
val = elem[1] # value
try:
chunk_arr[coord] = val # update the point
except IndexError:
msg = "Out of bounds point index for POST Chunk"
log.warn(msg)
raise HTTPBadRequest(reason=msg)
chunk_cache = app["chunk_cache"]
chunk_cache.setDirty(chunk_id)
# async write to S3
dirty_ids = app["dirty_ids"]
now = int(time.time())
dirty_ids[chunk_id] = (now, bucket)
log.info(f"set {chunk_id} to dirty")
else:
# reading point data
point_dt = np.dtype('uint64') # use unsigned long for point index
point_arr = np.fromstring(
input_bytes, dtype=point_dt) # read points as unsigned longs
if len(point_arr) % rank != 0:
msg = "Unexpected size of point array"
log.warn(msg)
raise HTTPBadRequest(reason=msg)
num_points = len(point_arr) // rank
log.debug(f"got {num_points} points")
point_arr = point_arr.reshape((num_points, rank))
output_arr = np.zeros((num_points, ), dtype=dset_dtype)
for i in range(num_points):
point = point_arr[i, :]
tr_point = getChunkRelativePoint(chunk_coord, point)
val = chunk_arr[tuple(tr_point)]
output_arr[i] = val
if put_points:
# write empty response
resp = json_response({})
else:
# get data
output_data = output_arr.tobytes()
# write response
try:
resp = StreamResponse()
resp.headers['Content-Type'] = "application/octet-stream"
resp.content_length = len(output_data)
await resp.prepare(request)
await resp.write(output_data)
except Exception as e:
log.error(f"Exception during binary data write: {e}")
raise HTTPInternalServerError()
finally:
await resp.write_eof()
return resp
async def PUT_Chunk(request):
log.request(request)
app = request.app
params = request.rel_url.query
chunk_id = request.match_info.get('id')
if not chunk_id:
msg = "Missing chunk id"
log.error(msg)
raise HTTPBadRequest(reason=msg)
if not isValidUuid(chunk_id, "Chunk"):
msg = "Invalid chunk id: {}".format(chunk_id)
log.warn(msg)
raise HTTPBadRequest(reason=msg)
if not request.has_body:
msg = "PUT Value with no body"
log.warn(msg)
raise HTTPBadRequest(reason=msg)
content_type = "application/octet-stream"
if "Content-Type" in request.headers:
# client should use "application/octet-stream" for binary transfer
content_type = request.headers["Content-Type"]
if content_type != "application/octet-stream":
msg = "Unexpected content_type: {}".format(content_type)
log.error(msg)
raise HTTPBadRequest(reason=msg)
validateInPartition(app, chunk_id)
log.debug("request params: {}".format(list(params.keys())))
if "dset" not in params:
msg = "Missing dset in GET request"
log.error(msg)
raise HTTPBadRequest(reason=msg)
dset_json = json.loads(params["dset"])
log.debug("dset_json: {}".format(dset_json))
dims = getChunkLayout(dset_json)
if "root" not in dset_json:
msg = "expected root key in dset_json"
log.error(msg)
raise KeyError(msg)
rank = len(dims)
# get chunk selection from query params
selection = []
for i in range(rank):
dim_slice = getSliceQueryParam(request, i, dims[i])
selection.append(dim_slice)
selection = tuple(selection)
log.debug("got selection: {}".format(selection))
type_json = dset_json["type"]
itemsize = 'H5T_VARIABLE'
if "size" in type_json:
itemsize = type_json["size"]
dt = createDataType(type_json)
log.debug("dtype: {}".format(dt))
if rank == 0:
msg = "No dimension passed to PUT chunk request"
log.error(msg)
raise HTTPBadRequest(reason=msg)
if len(selection) != rank:
msg = "Selection rank does not match shape rank"
log.error(msg)
raise HTTPBadRequest(reason=msg)
for i in range(rank):
s = selection[i]
log.debug("selection[{}]: {}".format(i, s))
mshape = getSelectionShape(selection)
log.debug(f"mshape: {mshape}")
num_elements = 1
for extent in mshape:
num_elements *= extent
# check that the content_length is what we expect
if itemsize != 'H5T_VARIABLE':
log.debug("expect content_length: {}".format(num_elements*itemsize))
log.debug("actual content_length: {}".format(request.content_length))
if itemsize != 'H5T_VARIABLE' and (num_elements * itemsize) != request.content_length:
msg = "Expected content_length of: {}, but got: {}".format(num_elements*itemsize, request.content_length)
log.error(msg)
raise HTTPBadRequest(reason=msg)
# create a numpy array for incoming data
input_bytes = await request_read(request) # TBD - will it cause problems when failures are raised before reading data?
if len(input_bytes) != request.content_length:
msg = "Read {} bytes, expecting: {}".format(len(input_bytes), request.content_length)
log.error(msg)
raise HTTPInternalServerError()
input_arr = bytesToArray(input_bytes, dt, mshape)
chunk_arr = await getChunk(app, chunk_id, dset_json, chunk_init=True)
# update chunk array
chunk_arr[selection] = input_arr
chunk_cache = app["chunk_cache"]
chunk_cache.setDirty(chunk_id)
log.info(f"PUT_Chunk dirty cache count: {chunk_cache.dirtyCount}")
# async write to S3
dirty_ids = app["dirty_ids"]
now = int(time.time())
dirty_ids[chunk_id] = now
# chunk update successful
resp = json_response({}, status=201)
log.response(request, resp=resp)
return resp
async def POST_Dataset(request):
"""HTTP method to create a new dataset object"""
log.request(request)
app = request.app
username, pswd = getUserPasswordFromRequest(request)
# write actions need auth
await validateUserPassword(app, username, pswd)
if not request.has_body:
msg = "POST Datasets with no body"
log.warn(msg)
raise HTTPBadRequest(reason=msg)
body = await request.json()
# get domain, check authorization
domain = getDomainFromRequest(request)
if not isValidDomain(domain):
msg = "Invalid host value: {}".format(domain)
log.warn(msg)
raise HTTPBadRequest(reason=msg)
domain_json = await getDomainJson(app, domain, reload=True)
root_id = domain_json["root"]
aclCheck(domain_json, "create",
username) # throws exception if not allowed
if "root" not in domain_json:
msg = "Expected root key for domain: {}".format(domain)
log.warn(msg)
raise HTTPBadRequest(reason=msg)
#
# validate type input
#
if "type" not in body:
msg = "POST Dataset has no type key in body"
log.warn(msg)
raise HTTPBadRequest(reason=msg)
datatype = body["type"]
if isinstance(datatype, str) and datatype.startswith("t-"):
# Committed type - fetch type json from DN
ctype_id = datatype
log.debug("got ctypeid: {}".format(ctype_id))
ctype_json = await getObjectJson(app, ctype_id)
log.debug("ctype: {}".format(ctype_json))
if ctype_json["root"] != root_id:
msg = "Referenced committed datatype must belong in same domain"
log.warn(msg)
raise HTTPBadRequest(reason=msg)
datatype = ctype_json["type"]
# add the ctype_id to type type
datatype["id"] = ctype_id
elif isinstance(datatype, str):
try:
# convert predefined type string (e.g. "H5T_STD_I32LE") to
# corresponding json representation
datatype = getBaseTypeJson(datatype)
log.debug("got datatype: {}".format(datatype))
except TypeError:
msg = "POST Dataset with invalid predefined type"
log.warn(msg)
raise HTTPBadRequest(reason=msg)
validateTypeItem(datatype)
item_size = getItemSize(datatype)
#
# Validate shape input
#
dims = None
shape_json = {}
if "shape" not in body:
shape_json["class"] = "H5S_SCALAR"
else:
shape = body["shape"]
if isinstance(shape, int):
shape_json["class"] = "H5S_SIMPLE"
dims = [
shape,
]
shape_json["dims"] = dims
elif isinstance(shape, str):
# only valid string value is H5S_NULL
if shape != "H5S_NULL":
msg = "POST Datset with invalid shape value"
log.warn(msg)
raise HTTPBadRequest(reason=msg)
shape_json["class"] = "H5S_NULL"
elif isinstance(shape, list):
if len(shape) == 0:
shape_json["class"] = "H5S_SCALAR"
else:
shape_json["class"] = "H5S_SIMPLE"
shape_json["dims"] = shape
dims = shape
else:
msg = "Bad Request: shape is invalid"
log.warn(msg)
raise HTTPBadRequest(reason=msg)
if dims is not None:
for i in range(len(dims)):
extent = dims[i]
if not isinstance(extent, int):
msg = "Invalid shape type"
log.warn(msg)
raise HTTPBadRequest(reason=msg)
if extent < 0:
msg = "shape dimension is negative"
log.warn(msg)
raise HTTPBadRequest(reason=msg)
maxdims = None
if "maxdims" in body:
if dims is None:
msg = "Maxdims cannot be supplied if space is NULL"
log.warn(msg)
raise HTTPBadRequest(reason=msg)
maxdims = body["maxdims"]
if isinstance(maxdims, int):
dim1 = maxdims
maxdims = [dim1]
elif isinstance(maxdims, list):
pass # can use as is
else:
msg = "Bad Request: maxdims is invalid"
log.warn(msg)
raise HTTPBadRequest(reason=msg)
if len(dims) != len(maxdims):
msg = "Maxdims rank doesn't match Shape"
log.warn(msg)
raise HTTPBadRequest(reason=msg)
if maxdims is not None:
for extent in maxdims:
if not isinstance(extent, int):
msg = "Invalid maxdims type"
log.warn(msg)
raise HTTPBadRequest(reason=msg)
if extent < 0:
msg = "maxdims dimension is negative"
log.warn(msg)
raise HTTPBadRequest(reason=msg)
if len(maxdims) != len(dims):
msg = "Bad Request: maxdims array length must equal shape array length"
log.warn(msg)
raise HTTPBadRequest(reason=msg)
shape_json["maxdims"] = []
for i in range(len(dims)):
maxextent = maxdims[i]
if not isinstance(maxextent, int):
msg = "Bad Request: maxdims must be integer type"
log.warn(msg)
raise HTTPBadRequest(reason=msg)
elif maxextent == 0:
# unlimited dimension
shape_json["maxdims"].append(0)
elif maxextent < dims[i]:
msg = "Bad Request: maxdims extent can't be smaller than shape extent"
log.warn(msg)
raise HTTPBadRequest(reason=msg)
else:
shape_json["maxdims"].append(maxextent)
layout = None
min_chunk_size = int(config.get("min_chunk_size"))
max_chunk_size = int(config.get("max_chunk_size"))
if 'creationProperties' in body:
creationProperties = body["creationProperties"]
if 'layout' in creationProperties:
layout = creationProperties["layout"]
validateChunkLayout(shape_json, item_size, layout)
if layout is None and shape_json["class"] != "H5S_NULL":
# default to chunked layout
layout = {"class": "H5D_CHUNKED"}
if layout and layout["class"] == 'H5D_CONTIGUOUS_REF':
chunk_dims = getContiguousLayout(shape_json,
item_size,
chunk_min=min_chunk_size,
chunk_max=max_chunk_size)
layout["dims"] = chunk_dims
log.debug(f"autoContiguous layout: {layout}")
if layout and layout["class"] == 'H5D_CHUNKED' and "dims" not in layout:
# do autochunking
chunk_dims = guessChunk(shape_json, item_size)
layout["dims"] = chunk_dims
log.debug(f"initial autochunk layout: {layout}")
if layout and layout["class"] == 'H5D_CHUNKED':
chunk_dims = layout["dims"]
chunk_size = getChunkSize(chunk_dims, item_size)
log.debug("chunk_size: {}, min: {}, max: {}".format(
chunk_size, min_chunk_size, max_chunk_size))
# adjust the chunk shape if chunk size is too small or too big
adjusted_chunk_dims = None
if chunk_size < min_chunk_size:
log.debug(
"chunk size: {} less than min size: {}, expanding".format(
chunk_size, min_chunk_size))
adjusted_chunk_dims = expandChunk(chunk_dims,
item_size,
shape_json,
chunk_min=min_chunk_size,
layout_class=layout["class"])
elif chunk_size > max_chunk_size:
log.debug(
"chunk size: {} greater than max size: {}, expanding".format(
chunk_size, max_chunk_size, layout_class=layout["class"]))
adjusted_chunk_dims = shrinkChunk(chunk_dims,
item_size,
chunk_max=max_chunk_size)
if adjusted_chunk_dims:
log.debug(
f"requested chunk_dimensions: {chunk_dims} modified dimensions: {adjusted_chunk_dims}"
)
layout["dims"] = adjusted_chunk_dims
if layout and layout["class"] in ('H5D_CHUNKED_REF',
'H5D_CHUNKED_REF_INDIRECT'):
chunk_dims = layout["dims"]
chunk_size = getChunkSize(chunk_dims, item_size)
log.debug("chunk_size: {}, min: {}, max: {}".format(
chunk_size, min_chunk_size, max_chunk_size))
# adjust the chunk shape if chunk size is too small or too big
if chunk_size < min_chunk_size:
log.warn(
"chunk size: {} less than min size: {} for H5D_CHUNKED_REF dataset"
.format(chunk_size, min_chunk_size))
elif chunk_size > max_chunk_size:
log.warn(
"chunk size: {} greater than max size: {}, for H5D_CHUNKED_REF dataset"
.format(chunk_size,
max_chunk_size,
layout_class=layout["class"]))
link_id = None
link_title = None
if "link" in body:
link_body = body["link"]
if "id" in link_body:
link_id = link_body["id"]
if "name" in link_body:
link_title = link_body["name"]
if link_id and link_title:
log.info("link id: {}".format(link_id))
# verify that the referenced id exists and is in this domain
# and that the requestor has permissions to create a link
await validateAction(app, domain, link_id, username, "create")
dset_id = createObjId("datasets", rootid=root_id)
log.info("new dataset id: {}".format(dset_id))
dataset_json = {
"id": dset_id,
"root": root_id,
"type": datatype,
"shape": shape_json
}
if "creationProperties" in body:
# TBD - validate all creationProperties
creationProperties = body["creationProperties"]
if "fillValue" in creationProperties:
# validate fill value compatible with type
dt = createDataType(datatype)
fill_value = creationProperties["fillValue"]
if isinstance(fill_value, list):
fill_value = tuple(fill_value)
try:
np.asarray(fill_value, dtype=dt)
except (TypeError, ValueError):
msg = "Fill value {} not compatible with dataset type: {}".format(
fill_value, datatype)
log.warn(msg)
raise HTTPBadRequest(reason=msg)
dataset_json["creationProperties"] = creationProperties
if layout is not None:
dataset_json["layout"] = layout
log.debug("create dataset: " + json.dumps(dataset_json))
req = getDataNodeUrl(app, dset_id) + "/datasets"
post_json = await http_post(app, req, data=dataset_json)
# create link if requested
if link_id and link_title:
link_json = {}
link_json["id"] = dset_id
link_json["class"] = "H5L_TYPE_HARD"
link_req = getDataNodeUrl(app, link_id)
link_req += "/groups/" + link_id + "/links/" + link_title
log.info("PUT link - : " + link_req)
put_rsp = await http_put(app, link_req, data=link_json)
log.debug("PUT Link resp: {}".format(put_rsp))
# dataset creation successful
resp = await jsonResponse(request, post_json, status=201)
log.response(request, resp=resp)
return resp
async def PUT_Chunk(request):
log.request(request)
app = request.app
params = request.rel_url.query
query = None
if "query" in params:
query = params["query"]
log.info(f"PUT_Chunk query: {query}")
chunk_id = request.match_info.get('id')
if not chunk_id:
msg = "Missing chunk id"
log.error(msg)
raise HTTPBadRequest(reason=msg)
if not isValidUuid(chunk_id, "Chunk"):
msg = f"Invalid chunk id: {chunk_id}"
log.warn(msg)
raise HTTPBadRequest(reason=msg)
if not request.has_body:
msg = "PUT Value with no body"
log.warn(msg)
raise HTTPBadRequest(reason=msg)
if "bucket" in params:
bucket = params["bucket"]
log.debug(f"PUT_Chunk using bucket: {bucket}")
else:
bucket = None
if query:
expected_content_type = "text/plain; charset=utf-8"
else:
expected_content_type = "application/octet-stream"
if "Content-Type" in request.headers:
# client should use "application/octet-stream" for binary transfer
content_type = request.headers["Content-Type"]
if content_type != expected_content_type:
msg = f"Unexpected content_type: {content_type}"
log.error(msg)
raise HTTPBadRequest(reason=msg)
validateInPartition(app, chunk_id)
if "dset" in params:
msg = "Unexpected param dset in GET request"
log.error(msg)
raise HTTPBadRequest(reason=msg)
log.debug(f"PUT_Chunk - id: {chunk_id}")
dset_id = getDatasetId(chunk_id)
dset_json = await get_metadata_obj(app, dset_id, bucket=bucket)
log.debug(f"dset_json: {dset_json}")
dims = getChunkLayout(dset_json)
if "root" not in dset_json:
msg = "expected root key in dset_json"
log.error(msg)
raise KeyError(msg)
rank = len(dims)
# get chunk selection from query params
selection = []
for i in range(rank):
dim_slice = getSliceQueryParam(request, i, dims[i])
selection.append(dim_slice)
selection = tuple(selection)
log.debug(f"got selection: {selection}")
type_json = dset_json["type"]
itemsize = 'H5T_VARIABLE'
if "size" in type_json:
itemsize = type_json["size"]
dt = createDataType(type_json)
log.debug(f"dtype: {dt}")
if rank == 0:
msg = "No dimension passed to PUT chunk request"
log.error(msg)
raise HTTPBadRequest(reason=msg)
if len(selection) != rank:
msg = "Selection rank does not match shape rank"
log.error(msg)
raise HTTPBadRequest(reason=msg)
for i in range(rank):
s = selection[i]
log.debug(f"selection[{i}]: {s}")
mshape = getSelectionShape(selection)
log.debug(f"mshape: {mshape}")
num_elements = 1
for extent in mshape:
num_elements *= extent
resp = {}
query_update = None
limit = 0
chunk_init = True
input_arr = None
if query:
if not dt.fields:
log.error("expected compound dtype for PUT query")
raise HTTPInternalServerError()
query_update = await request.json()
log.debug(f"query_update: {query_update}")
if "Limit" in params:
limit = int(params["Limit"])
chunk_init = False
else:
# regular chunk update
# check that the content_length is what we expect
if itemsize != 'H5T_VARIABLE':
log.debug(f"expect content_length: {num_elements*itemsize}")
log.debug(f"actual content_length: {request.content_length}")
if itemsize != 'H5T_VARIABLE' and (num_elements *
itemsize) != request.content_length:
msg = f"Expected content_length of: {num_elements*itemsize}, but got: {request.content_length}"
log.error(msg)
raise HTTPBadRequest(reason=msg)
# create a numpy array for incoming data
input_bytes = await request_read(
request
) # TBD - will it cause problems when failures are raised before reading data?
if len(input_bytes) != request.content_length:
msg = f"Read {len(input_bytes)} bytes, expecting: {request.content_length}"
log.error(msg)
raise HTTPInternalServerError()
input_arr = bytesToArray(input_bytes, dt, mshape)
# TBD: Skip read if the input shape is the entire chunk?
chunk_arr = await getChunk(app,
chunk_id,
dset_json,
chunk_init=chunk_init,
bucket=bucket)
is_dirty = False
if query:
values = []
indices = []
if chunk_arr is not None:
# do query selection
limit = 0
if "Limit" in params:
limit = int(params["Limit"])
field_names = list(dt.fields.keys())
replace_mask = [
None,
] * len(field_names)
for i in range(len(field_names)):
field_name = field_names[i]
if field_name in query_update:
replace_mask[i] = query_update[field_name]
log.debug(f"replace_mask: {replace_mask}")
x = chunk_arr[selection]
log.debug(f"put_query - x: {x}")
eval_str = getEvalStr(query, "x", field_names)
log.debug(f"put_query - eval_str: {eval_str}")
where_result = np.where(eval(eval_str))
log.debug(f"put_query - where_result: {where_result}")
where_result_index = where_result[0]
log.debug(f"put_query - whare_result index: {where_result_index}")
log.debug(
f"put_query - boolean selection: {x[where_result_index]}")
s = selection[0]
count = 0
for index in where_result_index:
log.debug(f"put_query - index: {index}")
value = x[index]
log.debug(f"put_query - original value: {value}")
for i in range(len(field_names)):
if replace_mask[i] is not None:
value[i] = replace_mask[i]
log.debug(f"put_query - modified value: {value}")
x[index] = value
json_val = bytesArrayToList(value)
log.debug(f"put_query - json_value: {json_val}")
json_index = index.tolist(
) * s.step + s.start # adjust for selection
indices.append(json_index)
values.append(json_val)
count += 1
is_dirty = True
if limit > 0 and count >= limit:
log.info("put_query - got limit items")
break
query_result = {}
query_result["index"] = indices
query_result["value"] = values
log.info(f"query_result retiurning: {len(indices)} rows")
log.debug(f"query_result: {query_result}")
resp = json_response(query_result)
else:
# update chunk array
chunk_arr[selection] = input_arr
is_dirty = True
resp = json_response({}, status=201)
if is_dirty:
chunk_cache = app["chunk_cache"]
chunk_cache.setDirty(chunk_id)
log.info(f"PUT_Chunk dirty cache count: {chunk_cache.dirtyCount}")
# async write to S3
dirty_ids = app["dirty_ids"]
now = int(time.time())
dirty_ids[chunk_id] = (now, bucket)
# chunk update successful
log.response(request, resp=resp)
return resp
async def PUT_Chunk(request):
log.request(request)
app = request.app
params = request.rel_url.query
query = None
query_update = None
limit = 0
bucket = None
input_arr = None
if "query" in params:
query = params["query"]
log.info(f"PUT_Chunk query: {query}")
if "Limit" in params:
limit = int(params["Limit"])
chunk_id = request.match_info.get('id')
if not chunk_id:
msg = "Missing chunk id"
log.error(msg)
raise HTTPBadRequest(reason=msg)
if not isValidUuid(chunk_id, "Chunk"):
msg = f"Invalid chunk id: {chunk_id}"
log.warn(msg)
raise HTTPBadRequest(reason=msg)
if not request.has_body:
msg = "PUT Value with no body"
log.warn(msg)
raise HTTPBadRequest(reason=msg)
if "bucket" in params:
bucket = params["bucket"]
log.debug(f"PUT_Chunk using bucket: {bucket}")
else:
bucket = None
if query:
expected_content_type = "text/plain; charset=utf-8"
chunk_init = False # don't initalize new chunks on query update
else:
expected_content_type = "application/octet-stream"
chunk_init = True
if "Content-Type" in request.headers:
# client should use "application/octet-stream" for binary transfer
content_type = request.headers["Content-Type"]
if content_type != expected_content_type:
msg = f"Unexpected content_type: {content_type}"
log.error(msg)
raise HTTPBadRequest(reason=msg)
validateInPartition(app, chunk_id)
if "dset" in params:
msg = "Unexpected param dset in GET request"
log.error(msg)
raise HTTPBadRequest(reason=msg)
log.debug(f"PUT_Chunk - id: {chunk_id}")
dset_id = getDatasetId(chunk_id)
dset_json = await get_metadata_obj(app, dset_id, bucket=bucket)
log.debug(f"dset_json: {dset_json}")
# TBD - does this work with linked datasets?
dims = getChunkLayout(dset_json)
log.debug(f"got dims: {dims}")
rank = len(dims)
type_json = dset_json["type"]
dt = createDataType(type_json)
log.debug(f"dtype: {dt}")
itemsize = 'H5T_VARIABLE'
if "size" in type_json:
itemsize = type_json["size"]
# get chunk selection from query params
selection = []
for i in range(rank):
dim_slice = getSliceQueryParam(request, i, dims[i])
selection.append(dim_slice)
selection = tuple(selection)
log.debug(f"got selection: {selection}")
mshape = getSelectionShape(selection)
log.debug(f"mshape: {mshape}")
num_elements = 1
for extent in mshape:
num_elements *= extent
chunk_arr = await get_chunk(app, chunk_id, dset_json, bucket=bucket, chunk_init=chunk_init)
is_dirty = False
if chunk_arr is None:
if chunk_init:
log.error(f"failed to create numpy array")
raise HTTPInternalServerError()
else:
log.warn(f"chunk {chunk_id} not found")
raise HTTPNotFound()
if query:
if not dt.fields:
log.error("expected compound dtype for PUT query")
raise HTTPInternalServerError()
if rank != 1:
log.error("expected one-dimensional array for PUT query")
raise HTTPInternalServerError()
query_update = await request.json()
log.debug(f"query_update: {query_update}")
# TBD - send back binary response to SN node
try:
resp = chunkQuery(chunk_id=chunk_id, chunk_layout=dims, chunk_arr=chunk_arr, slices=selection,
query=query, query_update=query_update, limit=limit, return_json=True)
except TypeError as te:
log.warn(f"chunkQuery - TypeError: {te}")
raise HTTPBadRequest()
except ValueError as ve:
log.warn(f"chunkQuery - ValueError: {ve}")
raise HTTPBadRequest()
if query_update and resp is not None:
is_dirty = True
else:
# regular chunk update
# check that the content_length is what we expect
if itemsize != 'H5T_VARIABLE':
log.debug(f"expect content_length: {num_elements*itemsize}")
log.debug(f"actual content_length: {request.content_length}")
if itemsize != 'H5T_VARIABLE' and (num_elements * itemsize) != request.content_length:
msg = f"Expected content_length of: {num_elements*itemsize}, but got: {request.content_length}"
log.error(msg)
raise HTTPBadRequest(reason=msg)
# create a numpy array for incoming data
input_bytes = await request_read(request) # TBD - will it cause problems when failures are raised before reading data?
if len(input_bytes) != request.content_length:
msg = f"Read {len(input_bytes)} bytes, expecting: {request.content_length}"
log.error(msg)
raise HTTPInternalServerError()
input_arr = bytesToArray(input_bytes, dt, mshape)
chunkWriteSelection(chunk_arr=chunk_arr, slices=selection, data=input_arr)
is_dirty = True
# chunk update successful
resp = {}
if is_dirty:
save_chunk(app, chunk_id, bucket=bucket)
resp = json_response(resp, status=201)
log.response(request, resp=resp)
return resp
async def POST_Chunk(request):
log.request(request)
app = request.app
params = request.rel_url.query
put_points = False
num_points = 0
if "count" not in params:
log.warn("expected count param")
raise HTTPBadRequest()
if "count" in params:
num_points = int(params["count"])
if "action" in params and params["action"] == "put":
log.info(f"POST Chunk put points - num_points: {num_points}")
put_points = True
else:
log.info(f"POST Chunk get points - num_points: {num_points}")
s3path = None
s3offset = 0
s3size = 0
if "s3path" in params:
if put_points:
log.error("s3path can not be used with put points POST request")
raise HTTPBadRequest()
s3path = params["s3path"]
log.debug(f"GET_Chunk - using s3path: {s3path}")
bucket = None
elif "bucket" in params:
bucket = params["bucket"]
else:
bucket = None
if "s3offset" in params:
try:
s3offset = int(params["s3offset"])
except ValueError:
log.error(f"invalid s3offset params: {params['s3offset']}")
raise HTTPBadRequest()
if "s3size" in params:
try:
s3size = int(params["s3size"])
except ValueError:
log.error(f"invalid s3size params: {params['s3sieze']}")
raise HTTPBadRequest()
chunk_id = request.match_info.get('id')
if not chunk_id:
msg = "Missing chunk id"
log.error(msg)
raise HTTPBadRequest(reason=msg)
log.info(f"POST chunk_id: {chunk_id}")
chunk_index = getChunkIndex(chunk_id)
log.debug(f"chunk_index: {chunk_index}")
if not isValidUuid(chunk_id, "Chunk"):
msg = f"Invalid chunk id: {chunk_id}"
log.warn(msg)
raise HTTPBadRequest(reason=msg)
validateInPartition(app, chunk_id)
log.debug(f"request params: {list(params.keys())}")
if "dset" in params:
msg = "Unexpected dset in POST request"
log.error(msg)
raise HTTPBadRequest(reason=msg)
if not request.has_body:
msg = "POST Value with no body"
log.warn(msg)
raise HTTPBadRequest(reason=msg)
content_type = "application/octet-stream"
if "Content-Type" in request.headers:
# client should use "application/octet-stream" for binary transfer
content_type = request.headers["Content-Type"]
if content_type != "application/octet-stream":
msg = f"Unexpected content_type: {content_type}"
log.error(msg)
raise HTTPBadRequest(reason=msg)
dset_id = getDatasetId(chunk_id)
dset_json = await get_metadata_obj(app, dset_id, bucket=bucket)
dims = getChunkLayout(dset_json)
rank = len(dims)
type_json = dset_json["type"]
dset_dtype = createDataType(type_json)
# create a numpy array for incoming points
input_bytes = await request_read(request)
if len(input_bytes) != request.content_length:
msg = f"Read {len(input_bytes)} bytes, expecting: {request.content_length}"
log.error(msg)
raise HTTPInternalServerError()
if rank == 1:
coord_type_str = "uint64"
else:
coord_type_str = f"({rank},)uint64"
if put_points:
# create a numpy array with the following type:
# (coord1, coord2, ...) | dset_dtype
point_dt = np.dtype([("coord", np.dtype(coord_type_str)), ("value", dset_dtype)])
point_shape = (num_points,)
chunk_init = True
else:
point_dt = np.dtype('uint64')
point_shape = (num_points, rank)
chunk_init = False
point_arr = bytesToArray(input_bytes, point_dt, point_shape)
chunk_arr = await get_chunk(app, chunk_id, dset_json, bucket=bucket, s3path=s3path, s3offset=s3offset, s3size=s3size, chunk_init=chunk_init)
if chunk_arr is None:
log.warn(f"chunk {chunk_id} not found")
raise HTTPNotFound()
if put_points:
# writing point data
try:
chunkWritePoints(chunk_id=chunk_id, chunk_layout=dims, chunk_arr=chunk_arr, point_arr=point_arr)
except ValueError as ve:
log.warn(f"got value error from chunkWritePoints: {ve}")
raise HTTPBadRequest()
# write empty response
resp = json_response({})
save_chunk(app, chunk_id, bucket=bucket) # lazily write chunk to storage
else:
# read points
try:
output_arr = chunkReadPoints(chunk_id=chunk_id, chunk_layout=dims, chunk_arr=chunk_arr, point_arr=point_arr)
except ValueError as ve:
log.warn(f"got value error from chunkReadPoints: {ve}")
raise HTTPBadRequest()
output_data = arrayToBytes(output_arr)
# write response
try:
resp = StreamResponse()
resp.headers['Content-Type'] = "application/octet-stream"
resp.content_length = len(output_data)
await resp.prepare(request)
await resp.write(output_data)
except Exception as e:
log.error(f"Exception during binary data write: {e}")
raise HTTPInternalServerError()
finally:
await resp.write_eof()
return resp
async def PUT_AttributeValue(request):
"""HTTP method to update an attributes data"""
log.request(request)
log.info("PUT_AttributeValue")
app = request.app
collection = getRequestCollectionName(request) # returns datasets|groups|datatypes
obj_id = request.match_info.get('id')
if not obj_id:
msg = "Missing object id"
log.warn(msg)
raise HTTPBadRequest(reason=msg)
if not isValidUuid(obj_id, obj_class=collection):
msg = f"Invalid object id: {obj_id}"
log.warn(msg)
raise HTTPBadRequest(reason=msg)
attr_name = request.match_info.get('name')
log.debug(f"Attribute name: [{attr_name}]")
validateAttributeName(attr_name)
log.info(f"PUT Attribute Value id: {obj_id} name: {attr_name}")
username, pswd = getUserPasswordFromRequest(request)
# write actions need auth
await validateUserPassword(app, username, pswd)
if not request.has_body:
msg = "PUT AttributeValue with no body"
log.warn(msg)
raise HTTPBadRequest(reason=msg)
domain = getDomainFromRequest(request)
if not isValidDomain(domain):
msg = f"Invalid domain: {domain}"
log.warn(msg)
raise HTTPBadRequest(reason=msg)
bucket = getBucketForDomain(domain)
# get domain JSON
domain_json = await getDomainJson(app, domain)
if "root" not in domain_json:
log.error(f"Expected root key for domain: {domain}")
raise HTTPInternalServerError()
# TBD - verify that the obj_id belongs to the given domain
await validateAction(app, domain, obj_id, username, "update")
req = getDataNodeUrl(app, obj_id)
req += '/' + collection + '/' + obj_id + "/attributes/" + attr_name
log.debug("get Attribute: " + req)
params = {}
if bucket:
params["bucket"] = bucket
dn_json = await http_get(app, req, params=params)
log.debug("got attributes json from dn for obj_id: " + str(obj_id))
log.debug(f"got dn_json: {dn_json}")
attr_shape = dn_json["shape"]
if attr_shape["class"] == 'H5S_NULL':
msg = "Null space attributes can not be updated"
log.warn(msg)
raise HTTPBadRequest(reason=msg)
np_shape = getShapeDims(attr_shape)
type_json = dn_json["type"]
np_dtype = createDataType(type_json) # np datatype
request_type = "json"
if "Content-Type" in request.headers:
# client should use "application/octet-stream" for binary transfer
content_type = request.headers["Content-Type"]
if content_type not in ("application/json", "application/octet-stream"):
msg = f"Unknown content_type: {content_type}"
log.warn(msg)
raise HTTPBadRequest(reason=msg)
if content_type == "application/octet-stream":
log.debug("PUT AttributeValue - request_type is binary")
request_type = "binary"
else:
log.debug("PUT AttribueValue - request type is json")
binary_data = None
if request_type == "binary":
item_size = getItemSize(type_json)
if item_size == 'H5T_VARIABLE':
msg = "Only JSON is supported for variable length data types"
log.warn(msg)
raise HTTPBadRequest(reason=msg)
# read binary data
binary_data = await request.read()
if len(binary_data) != request.content_length:
msg = f"Read {len(binary_data)} bytes, expecting: {request.content_length}"
log.error(msg)
raise HTTPInternalServerError()
arr = None # np array to hold request data
if binary_data:
npoints = getNumElements(np_shape)
if npoints*item_size != len(binary_data):
msg = "Expected: " + str(npoints*item_size) + " bytes, but got: " + str(len(binary_data))
log.warn(msg)
raise HTTPBadRequest(reason=msg)
arr = np.fromstring(binary_data, dtype=np_dtype)
arr = arr.reshape(np_shape) # conform to selection shape
# convert to JSON for transmission to DN
data = arr.tolist()
value = bytesArrayToList(data)
else:
body = await request.json()
if "value" not in body:
msg = "PUT attribute value with no value in body"
log.warn(msg)
raise HTTPBadRequest(reason=msg)
value = body["value"]
# validate that the value agrees with type/shape
try:
arr = jsonToArray(np_shape, np_dtype, value)
except ValueError:
msg = "Bad Request: input data doesn't match selection"
log.warn(msg)
raise HTTPBadRequest(reason=msg)
log.info(f"Got: {arr.size} array elements")
# ready to add attribute now
attr_json = {}
attr_json["type"] = type_json
attr_json["shape"] = attr_shape
attr_json["value"] = value
req = getDataNodeUrl(app, obj_id)
req += '/' + collection + '/' + obj_id + "/attributes/" + attr_name
log.info(f"PUT Attribute Value: {req}")
dn_json["value"] = value
params = {}
params = {"replace": 1} # let the DN know we can overwrite the attribute
if bucket:
params["bucket"] = bucket
put_rsp = await http_put(app, req, params=params, data=attr_json)
log.info(f"PUT Attribute Value resp: {put_rsp}")
hrefs = [] # TBD
req_rsp = { "hrefs": hrefs }
# attribute creation successful
resp = await jsonResponse(request, req_rsp)
log.response(request, resp=resp)
return resp
async def GET_AttributeValue(request):
"""HTTP method to return an attribute value"""
log.request(request)
app = request.app
log.info("GET_AttributeValue")
collection = getRequestCollectionName(request) # returns datasets|groups|datatypes
obj_id = request.match_info.get('id')
if not obj_id:
msg = "Missing object id"
log.warn(msg)
raise HTTPBadRequest(reason=msg)
if not isValidUuid(obj_id, obj_class=collection):
msg = f"Invalid object id: {obj_id}"
log.warn(msg)
raise HTTPBadRequest(reason=msg)
attr_name = request.match_info.get('name')
validateAttributeName(attr_name)
username, pswd = getUserPasswordFromRequest(request)
if username is None and app['allow_noauth']:
username = "******"
else:
await validateUserPassword(app, username, pswd)
domain = getDomainFromRequest(request)
if not isValidDomain(domain):
msg = f"Invalid domain value: {domain}"
log.warn(msg)
raise HTTPBadRequest(reason=msg)
bucket = getBucketForDomain(domain)
# get domain JSON
domain_json = await getDomainJson(app, domain)
if "root" not in domain_json:
log.error(f"Expected root key for domain: {domain}")
raise HTTPBadRequest(reason="Unexpected Error")
# TBD - verify that the obj_id belongs to the given domain
await validateAction(app, domain, obj_id, username, "read")
req = getDataNodeUrl(app, obj_id)
req += '/' + collection + '/' + obj_id + "/attributes/" + attr_name
log.debug("get Attribute: " + req)
params = {}
if bucket:
params["bucket"] = bucket
dn_json = await http_get(app, req, params=params)
log.debug("got attributes json from dn for obj_id: " + str(dn_json))
attr_shape = dn_json["shape"]
log.debug(f"attribute shape: {attr_shape}")
if attr_shape["class"] == 'H5S_NULL':
msg = "Null space attributes can not be read"
log.warn(msg)
raise HTTPBadRequest(reason=msg)
accept_type = getAcceptType(request)
response_type = accept_type # will adjust later if binary not possible
type_json = dn_json["type"]
shape_json = dn_json["shape"]
item_size = getItemSize(type_json)
if item_size == 'H5T_VARIABLE' and accept_type != "json":
msg = "Client requested binary, but only JSON is supported for variable length data types"
log.info(msg)
response_type = "json"
if response_type == "binary":
arr_dtype = createDataType(type_json) # np datatype
np_shape = getShapeDims(shape_json)
try:
arr = jsonToArray(np_shape, arr_dtype, dn_json["value"])
except ValueError:
msg = "Bad Request: input data doesn't match selection"
log.warn(msg)
raise HTTPBadRequest(reason=msg)
output_data = arr.tobytes()
log.debug(f"GET AttributeValue - returning {len(output_data)} bytes binary data")
# write response
try:
resp = StreamResponse()
resp.content_type = "application/octet-stream"
resp.content_length = len(output_data)
# allow CORS
resp.headers['Access-Control-Allow-Origin'] = '*'
resp.headers['Access-Control-Allow-Methods'] = "GET, POST, DELETE, PUT, OPTIONS"
resp.headers['Access-Control-Allow-Headers'] = "Content-Type, api_key, Authorization"
await resp.prepare(request)
await resp.write(output_data)
except Exception as e:
log.error(f"Got exception: {e}")
raise HTTPInternalServerError()
finally:
await resp.write_eof()
else:
resp_json = {}
if "value" in dn_json:
resp_json["value"] = dn_json["value"]
hrefs = []
obj_uri = '/' + collection + '/' + obj_id
attr_uri = obj_uri + '/attributes/' + attr_name
hrefs.append({'rel': 'self', 'href': getHref(request, attr_uri)})
hrefs.append({'rel': 'home', 'href': getHref(request, '/')})
hrefs.append({'rel': 'owner', 'href': getHref(request, obj_uri)})
resp_json["hrefs"] = hrefs
resp = await jsonResponse(request, resp_json)
log.response(request, resp=resp)
return resp
async def PUT_Attribute(request):
"""HTTP method to create a new attribute"""
log.request(request)
app = request.app
collection = getRequestCollectionName(request) # returns datasets|groups|datatypes
obj_id = request.match_info.get('id')
if not obj_id:
msg = "Missing object id"
log.warn(msg)
raise HTTPBadRequest(reason=msg)
if not isValidUuid(obj_id, obj_class=collection):
msg = f"Invalid object id: {obj_id}"
log.warn(msg)
raise HTTPBadRequest(reason=msg)
attr_name = request.match_info.get('name')
log.debug(f"Attribute name: [{attr_name}]")
validateAttributeName(attr_name)
log.info(f"PUT Attribute id: {obj_id} name: {attr_name}")
username, pswd = getUserPasswordFromRequest(request)
# write actions need auth
await validateUserPassword(app, username, pswd)
if not request.has_body:
msg = "PUT Attribute with no body"
log.warn(msg)
raise HTTPBadRequest(reason=msg)
body = await request.json()
domain = getDomainFromRequest(request)
if not isValidDomain(domain):
msg = f"Invalid domain: {domain}"
log.warn(msg)
raise HTTPBadRequest(reason=msg)
bucket = getBucketForDomain(domain)
# get domain JSON
domain_json = await getDomainJson(app, domain)
if "root" not in domain_json:
log.error(f"Expected root key for domain: {domain}")
raise HTTPBadRequest(reason="Unexpected Error")
root_id = domain_json["root"]
# TBD - verify that the obj_id belongs to the given domain
await validateAction(app, domain, obj_id, username, "create")
if "type" not in body:
msg = "PUT attribute with no type in body"
log.warn(msg)
raise HTTPBadRequest(reason=msg)
datatype = body["type"]
if isinstance(datatype, str) and datatype.startswith("t-"):
# Committed type - fetch type json from DN
ctype_id = datatype
log.debug(f"got ctypeid: {ctype_id}")
ctype_json = await getObjectJson(app, ctype_id, bucket=bucket)
log.debug(f"ctype {ctype_id}: {ctype_json}")
if ctype_json["root"] != root_id:
msg = "Referenced committed datatype must belong in same domain"
log.warn(msg)
raise HTTPBadRequest(reason=msg)
datatype = ctype_json["type"]
# add the ctype_id to type type
datatype["id"] = ctype_id
elif isinstance(datatype, str):
try:
# convert predefined type string (e.g. "H5T_STD_I32LE") to
# corresponding json representation
datatype = getBaseTypeJson(datatype)
log.debug(f"got datatype: {datatype}")
except TypeError:
msg = "PUT attribute with invalid predefined type"
log.warn(msg)
raise HTTPBadRequest(reason=msg)
validateTypeItem(datatype)
dims = None
shape_json = {}
if "shape" in body:
shape_body = body["shape"]
shape_class = None
if isinstance(shape_body, dict) and "class" in shape_body:
shape_class = shape_body["class"]
elif isinstance(shape_body, str):
shape_class = shape_body
if shape_class:
if shape_class == "H5S_NULL":
shape_json["class"] = "H5S_NULL"
if isinstance(shape_body, dict) and "dims" in shape_body:
msg = "can't include dims with null shape"
log.warn(msg)
raise HTTPBadRequest(reason=msg)
if isinstance(shape_body, dict) and "value" in body:
msg = "can't have H5S_NULL shape with value"
log.warn(msg)
raise HTTPBadRequest(reason=msg)
elif shape_class == "H5S_SCALAR":
shape_json["class"] = "H5S_SCALAR"
dims = getShapeDims(shape_body)
if len(dims) != 1 or dims[0] != 1:
msg = "dimensions aren't valid for scalar attribute"
log.warn(msg)
raise HTTPBadRequest(reason=msg)
elif shape_class == "H5S_SIMPLE":
shape_json["class"] = "H5S_SIMPLE"
dims = getShapeDims(shape_body)
shape_json["dims"] = dims
else:
msg = f"Unknown shape class: {shape_class}"
log.warn(msg)
raise HTTPBadRequest(reason=msg)
else:
# no class, interpet shape value as dimensions and
# use H5S_SIMPLE as class
if isinstance(shape_body, list) and len(shape_body) == 0:
shape_json["class"] = "H5S_SCALAR"
dims = [1,]
else:
shape_json["class"] = "H5S_SIMPLE"
dims = getShapeDims(shape_body)
shape_json["dims"] = dims
else:
shape_json["class"] = "H5S_SCALAR"
dims = [1,]
if "value" in body:
if dims is None:
msg = "Bad Request: data can not be included with H5S_NULL space"
log.warn(msg)
raise HTTPBadRequest(reason=msg)
value = body["value"]
# validate that the value agrees with type/shape
arr_dtype = createDataType(datatype) # np datatype
if len(dims) == 0:
np_dims = [1,]
else:
np_dims = dims
log.debug(f"attribute dims: {np_dims}")
log.debug(f"attribute value: {value}")
try:
arr = jsonToArray(np_dims, arr_dtype, value)
except ValueError:
msg = "Bad Request: input data doesn't match selection"
log.warn(msg)
raise HTTPBadRequest(reason=msg)
log.info(f"Got: {arr.size} array elements")
else:
value = None
# ready to add attribute now
req = getDataNodeUrl(app, obj_id)
req += '/' + collection + '/' + obj_id + "/attributes/" + attr_name
log.info("PUT Attribute: " + req)
attr_json = {}
attr_json["type"] = datatype
attr_json["shape"] = shape_json
if value is not None:
attr_json["value"] = value
params = {}
if bucket:
params["bucket"] = bucket
put_rsp = await http_put(app, req, params=params, data=attr_json)
log.info(f"PUT Attribute resp: {put_rsp}")
hrefs = [] # TBD
req_rsp = { "hrefs": hrefs }
# attribute creation successful
resp = await jsonResponse(request, req_rsp, status=201)
log.response(request, resp=resp)
return resp
