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 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 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 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 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 = f"Invalid chunk id: {chunk_id}"
log.warn(msg)
raise HTTPBadRequest(reason=msg)
validateInPartition(app, chunk_id)
log.debug(f"request params: {params.keys()}")
bucket = None
s3path = None
s3offset = None
s3size = None
query = None
limit = 0
if "s3path" in params:
s3path = params["s3path"]
log.debug(f"GET_Chunk - using s3path: {s3path}")
elif "bucket" in params:
bucket = params["bucket"]
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()
if "query" in params:
query = params["query"]
if "Limit" in params:
limit = int(params["Limit"])
dset_id = getDatasetId(chunk_id)
dset_json = await get_metadata_obj(app, dset_id, bucket=bucket)
dims = getChunkLayout(dset_json)
log.debug(f"got dims: {dims}")
rank = len(dims)
log.debug(f"dset_json: {dset_json}")
# 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}")
chunk_arr = await get_chunk(app, chunk_id, dset_json, bucket=bucket, s3path=s3path, s3offset=s3offset, s3size=s3size, chunk_init=False)
if chunk_arr is None:
msg = f"chunk {chunk_id} not found"
log.warn(msg)
raise HTTPNotFound()
if query:
# run given query
try:
read_resp = chunkQuery(chunk_id=chunk_id, chunk_layout=dims, chunk_arr=chunk_arr, slices=selection,
query=query, 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()
else:
# read selected data from chunk
output_arr = chunkReadSelection(chunk_arr, slices=selection)
read_resp = arrayToBytes(output_arr)
# write response
if isinstance(read_resp, bytes):
try:
resp = StreamResponse()
resp.headers['Content-Type'] = "application/octet-stream"
resp.content_length = len(read_resp)
await resp.prepare(request)
await resp.write(read_resp)
except Exception as e:
log.error(f"Exception during binary data write: {e}")
raise HTTPInternalServerError()
finally:
await resp.write_eof()
else:
# JSON response
resp = json_response(read_resp)
return resp