def test_encrypted_parquet_read_schema_no_decryption_config(
tempdir, data_table):
"""Write an encrypted parquet, verify it's encrypted,
but then try to read its schema without decryption properties."""
test_encrypted_parquet_write_read(tempdir, data_table)
with pytest.raises(IOError, match=r"no decryption"):
pq.read_schema(tempdir / PARQUET_NAME)
def get_filters(
parquet: BytesIO, condition_groups: List[ConditionGroup]
) -> Optional[List[PyArrowConditionGroup]]:
""" Given a dataframe and list of conditions, filter the dataframe using a boolean OR """
schema = pq.read_schema(parquet)
filters: List[PyArrowConditionGroup] = []
for group in condition_groups:
group_tuple = []
for entry in group["conditions"]:
try:
field = schema.field(entry["column"])
if field.type == pyarrow.null():
# Empty dataset; thus no inherent data type for entry["column"]
continue
group_tuple.append((
entry["column"],
entry["operator"],
entry["values"],
))
except KeyError:
# Column does not exist within schema
continue
if group_tuple:
filters.append(group_tuple)
return filters or None
def file_to_schema(input_tuple):
sname, fpath = input_tuple
try:
df = pq.read_schema(fpath)
except:
print("ERROR", fpath)
return ';'.join([str(df.names), str(df.types)])
def run_partition_test(input_file: str, output_dir: str, filters: Optional[list] = None):
milliseconds_since_epoch = int(time() * 1000)
print('Parquet metadata: ' + str(pq.read_metadata(input_file)))
print('Parquet schema: ' + pq.read_schema(input_file).to_string())
data = pq.read_table(source=input_file, filters=filters)
# Write a dataset and collect metadata information of all written files
metadata_collector = []
root_path = output_dir + 'partitioned_' + str(milliseconds_since_epoch)
pq.write_to_dataset(data,
root_path=root_path,
partition_cols=['start_year'],
metadata_collector=metadata_collector)
# Write the ``_common_metadata`` parquet file without row groups statistics
pq.write_metadata(data.schema, root_path + '/_common_metadata')
# Write the ``_metadata`` parquet file with row groups statistics of all files
# Gives following error:
# File "pyarrow/_parquet.pyx", line 616, in pyarrow._parquet.FileMetaData.append_row_groups
# RuntimeError: AppendRowGroups requires equal schemas.
# data.schema has one more column than partitioned files when partitioning by one column
# Related? https://github.com/dask/dask/issues/6243
# pq.write_metadata(data.schema, root_path + '/_metadata', metadata_collector=metadata_collector)
# Read from partitioned dataset
# use the new generic Dataset API
start_year = 2018
value = 50000
table = pq.read_table(root_path,
filters=[('start_year', '>=', start_year), ('value', '>', value)])
# filters=[('start_year', '>=', start_year)])
print(table.to_pandas())
def test_read_schema(tempdir):
N = 100
df = pd.DataFrame({
'index': np.arange(N),
'values': np.random.randn(N)
}, columns=['index', 'values'])
data_path = tempdir / 'test.parquet'
table = pa.Table.from_pandas(df)
_write_table(table, data_path)
read1 = pq.read_schema(data_path)
read2 = pq.read_schema(data_path, memory_map=True)
assert table.schema.equals(read1)
assert table.schema.equals(read2)
assert table.schema.metadata[b'pandas'] == read1.metadata[b'pandas']
def columns_from_parquet_file(input_parquet):
schema = pa_parquet.read_schema(input_parquet)
column_specifications = schema.pandas_metadata['columns']
options = []
for col_item in column_specifications:
name = col_item['name']
if name is not None:
options.append({'label': name, 'value': name})
return options
def test_write_parquet_no_index():
df = pd.DataFrame(np.random.randn(6, 4),
columns=list('abcd'),
index=np.arange(6))
with tempfile.TemporaryDirectory() as tmpdir:
fs = LocalFileSystem()
filename = os.path.join(tmpdir, 'df.parquet')
write_dataframe_as_parquet(df, fs, filename)
schema = parquet.read_schema(os.path.join(tmpdir, filename))
assert '__index_level_0__' not in schema.names
assert df.columns.values.tolist() == schema.names
def askUserInput(self):
filepath = self.userdata['filepath']
s = pa.read_schema(filepath)
colnames = s.names
def cb(columns):
self.userdata['columns'] = columns
dlgchoice = DlgListChoice(colnames, 'Open Parquet',
'Choose curves to load')
dlgchoice.dlgdata.connect(cb)
dlgchoice.exec_()
def _bytes2schema(data):
reader = pa.BufferReader(data)
schema = pq.read_schema(reader)
fields = []
for idx in range(len(schema)):
f = schema[idx]
# schema data recovered from parquet always contains timestamp data in us-granularity, but pandas will use
# ns-granularity, so we re-align the two different worlds here
if f.type == pa.timestamp("us"):
f = pa.field(f.name, pa.timestamp("ns"))
fields.append(f)
return pa.schema(fields, schema.metadata)
def run_test(input_file: str, output_dir: str, filters: list, use_pandas: bool):
print('Using pyarrow')
print('Parquet metadata: ' + str(pq.read_metadata(input_file)))
print('Parquet schema: ' + pq.read_schema(input_file).to_string())
pq_file = pq.ParquetFile(input_file)
row_group_0_metadata = pq_file.metadata.row_group(0)
print('Parquet min for column 0, row group 0: ' + str(row_group_0_metadata.column(0).statistics.min))
print('Parquet max for column 0, row group 0: ' + str(row_group_0_metadata.column(0).statistics.max))
if use_pandas:
unfiltered_pandas_data = pq.read_table(source=input_file).to_pandas()
size = sys.getsizeof(unfiltered_pandas_data)
print('Size of UN-filtered pandas DataFrame in memory: ' + str(size) + ' bytes (' + str(size / 1000000) + ' MB)')
with timeblock('pyarrow read and filter'):
data = pq.read_table(source=input_file, filters=filters)
size = sys.getsizeof(data)
print('Size of filtered pyarrow table in memory: ' + str(size) + ' bytes (' + str(size / 1000000) + ' MB)')
if use_pandas:
unfiltered_pandas_data = data.to_pandas()
size = sys.getsizeof(unfiltered_pandas_data)
print('Size of filtered pandas DataFrame in memory: ' + str(size) + ' bytes (' + str(size / 1000000) + ' MB)')
# print(pandas_data.head(10))
milliseconds_since_epoch = int(time() * 1000)
output_file = output_dir + str(milliseconds_since_epoch) + '.parquet'
print('Output file name: ' + output_file)
with timeblock('pyarrow write_table()'):
pq.write_table(data, output_file)
print('Parquet metadata of output: ' + str(pq.read_metadata(output_file)))
print('Parquet schema of output: ' + pq.read_schema(output_file).to_string())
print('Size of output file on disk: ' + str(os.path.getsize(output_file)) + ' bytes ('
+ str(os.path.getsize(output_file) / 1000000) + ' MB)')
def read_encrypted_parquet(path, decryption_config, kms_connection_config,
crypto_factory):
file_decryption_properties = crypto_factory.file_decryption_properties(
kms_connection_config, decryption_config)
assert (file_decryption_properties is not None)
meta = pq.read_metadata(path,
decryption_properties=file_decryption_properties)
assert (meta.num_columns == 3)
schema = pq.read_schema(path,
decryption_properties=file_decryption_properties)
assert (len(schema.names) == 3)
result = pq.ParquetFile(path,
decryption_properties=file_decryption_properties)
return result.read(use_threads=True)
def test_read_schema(tmpdir):
import pyarrow.parquet as pq
N = 100
df = pd.DataFrame({
'index': np.arange(N),
'values': np.random.randn(N)
}, columns=['index', 'values'])
data_path = pjoin(str(tmpdir), 'test.parquet')
table = pa.Table.from_pandas(df)
_write_table(table, data_path)
assert table.schema.equals(pq.read_schema(data_path))
def file_to_schema_bad(input_tuple):
sname, fpath = input_tuple
matched = False
#corrupted = False
try:
df = pq.read_schema(fpath)
string_match = ';'.join([str(df.names), str(df.types)])
if sname in proper_schema:
if proper_schema[sname] == string_match:
matched = True
return (matched, sname, fpath, proper_schema[sname], string_match,
"CORRECT-FILE")
except:
return (matched, sname, fpath, proper_schema[sname], "NULL", "NULL",
"CORRUPTED-FILE")
def transform_ts(start, end, file):
train_columns = pq.read_schema(
file).names # List with all column names to test
# print(train_columns)
X = pd.DataFrame(data=None)
for i in train_columns[start:end]:
df_signal = pq.read_pandas(file, columns=[i]).to_pandas()
# turn parquet to dataframe of one single signal
# print("Shape of signal data {}".format(df_signal.shape))
sig = np.ravel(df_signal.iloc[:, 0].to_numpy()) # turn to numpy
t = df_signal.index.to_numpy() # turn time to numpy
x_dn = de_noising(high_pass_filter(sig))
x_deleted = delete_repeat(x_dn)
x_deleted_cond = (x_deleted < 99998)
x_deleted = x_deleted[x_deleted_cond]
print(x_deleted.shape)
t_deleted = t[x_deleted_cond]
# Generating New Time Series Features from signal
master_train = pd.DataFrame({
0: x_deleted,
1: np.repeat(i, x_deleted.shape[0]),
2: t_deleted
})
# print("Shape of master train data {}".format(master_train.shape))
# master_train.to_csv('output/master_train.csv')
extraction_settings = EfficientFCParameters()
X_signal = extract_features(
master_train,
column_id=1,
column_sort=2,
impute_function=impute,
default_fc_parameters=extraction_settings)
print("Number of extracted features in {}: {}.".format(
i, X_signal.shape[1]))
X = X.append(X_signal)
return X
def run_id_filter_test(input_file: str, input_id_file: str):
# converting ids to pandas will be a "zero copy conversion" as unit_id column is int64 when:
# - ids are not nulls
# - a single ChunkedArray
# TODO check it that is the case
# https://arrow.apache.org/docs/python/pandas.html#zero-copy-series-conversions
filter_ids = pq.read_table(source=input_id_file)
filter_ids_as_pandas: DataFrame = filter_ids.to_pandas()
# filter_ids_as_list = filter_ids_as_pandas['unit_id'].tolist()
filter_ids_as_set = set(filter_ids_as_pandas['unit_id'])
print('Parquet metadata: ' + str(pq.read_metadata(input_id_file)))
print('Parquet schema: ' + pq.read_schema(input_id_file).to_string())
print('Using filter ids: ' + str(filter_ids.to_pandas()))
table = pq.read_table(source=input_file, filters=[
# ('unit_id', 'in', filter_ids_as_list)
('unit_id', 'in', filter_ids_as_set)
])
print(table.to_pandas())
writer = pq.ParquetWriter(outputFileName,schema) # ,coerce_timestamps='ms'
if args.debug is True:
totalWriteTime = totalWriteTime + (time.time() - startWrite)
if writer:
writer.close()
# print final success message.
timeNow = time.time()
elapsed = timeNow - timeStart
rps = int(rowcount / elapsed)
print (f"{rowcount} rows exported to {outputFileName} at {rps} rows per second.")
if args.debug is True:
print ()
print ('Performance breakdown:')
print (f'Read={totalReadTime:.2f} Transform={totalTransformTime:.2f} Write={totalWriteTime:.2f}')
print ()
print ('Test output display:')
print ()
print ('Schema:')
print (pq.read_schema(outputFileName))
print ()
testTable = pq.read_table (outputFileName)
print ('Data (10 rows):')
print (testTable.to_pandas().head(10))
exit(0)
def _InferArrowSchema(self):
match_result = FileSystems.match([self._file_pattern])[0]
files_metadata = match_result.metadata_list[0]
with FileSystems.open(files_metadata.path) as f:
return pq.read_schema(f)
def peek_metadata(path_to_table):
"""Read metadata without loading the data file
"""
schema = pq.read_schema(path_to_table)
return json.loads(schema.metadata[b'my_metadata'].decode('utf-8'))
try:
for stream in hdfs.ls(study):
try:
for version in hdfs.ls(stream):
try:
for user in hdfs.ls(version):
try:
files = hdfs.ls(user)
#if len(files)>2: put this check back again if schema mismatch is required
old_schema = []
for fle in hdfs.ls(user):
try:
if "_SUCCESS" not in fle:
with hdfs.open(hdfs_url +
fle) as f:
current_schema = pq.read_schema(
f).names
mismatched_metadata.append({
"total_files":
len(files),
"file_name":
fle,
"schema":
current_schema,
"user_folder":
user
})
except Exception as e:
print(str(e))
except Exception as e:
print(str(e))
except Exception as e:
def forced_extraction(sources_df: pd.DataFrame, cfg_err_ra: float,
cfg_err_dec: float, p_run: Run, extr_df: pd.DataFrame,
min_sigma: float, edge_buffer: float,
cluster_threshold: float, allow_nan: bool,
add_mode: bool, done_images_df: pd.DataFrame,
done_source_ids: List[int]) -> Tuple[pd.DataFrame, int]:
"""
Check and extract expected measurements, and associated them with the
related source(s).
Args:
sources_df:
Dataframe containing all the extracted measurements and
associations (product from association step).
cfg_err_ra:
The minimum RA error from the config file (in degrees).
cfg_err_dec:
The minimum declination error from the config file (in degrees).
p_run:
The pipeline run object.
extr_df:
The dataframe containing the information on what sources are
missing from which images (output from
get_src_skyregion_merged_df in main.py).
min_sigma:
Minimum sigma value to drop forced extracted measurements.
edge_buffer:
Flag to pass to ForcedPhot.measure method.
cluster_threshold:
Flag to pass to ForcedPhot.measure method.
allow_nan:
Flag to pass to ForcedPhot.measure method.
add_mode:
True when the pipeline is running in add image mode.
done_images_df:
Dataframe containing the images that thave already been processed
in a previous run (used in add image mode).
done_source_ids:
List of the source ids that were already present in the previous
run (used in add image mode).
Returns:
The sources_df with the extracted sources added and n_forced is the
total number of forced measurements present in the run.
"""
logger.info('Starting force extraction step.')
timer = StopWatch()
# get all the skyregions and related images
cols = [
'id', 'name', 'measurements_path', 'path', 'noise_path', 'beam_bmaj',
'beam_bmin', 'beam_bpa', 'background_path', 'rms_min', 'datetime',
'skyreg__centre_ra', 'skyreg__centre_dec', 'skyreg__xtr_radius'
]
images_df = pd.DataFrame(
list(
Image.objects.filter(run=p_run).select_related('skyreg').order_by(
'datetime').values(*tuple(cols)))).set_index('name')
# | name | id | measurements_path | path | noise_path |
# |:------------------------------|-----:|:--------------------|:-------------|:-------------|
# | VAST_2118-06A.EPOCH01.I.fits | 1 | path/to/file | path/to/file | path/to/file |
# | VAST_2118-06A.EPOCH03x.I.fits | 3 | path/to/file | path/to/file | path/to/file |
# | VAST_2118-06A.EPOCH02.I.fits | 2 | path/to/file | path/to/file | path/to/file |
# | name | beam_bmaj | beam_bmin | beam_bpa | background_path |
# |:------------------------------|------------:|------------:|-----------:|:------------------|
# | VAST_2118-06A.EPOCH01.I.fits | 0.00589921 | 0.00326088 | -70.4032 | path/to/file |
# | VAST_2118-06A.EPOCH03x.I.fits | 0.00470991 | 0.00300502 | -83.1128 | path/to/file |
# | VAST_2118-06A.EPOCH02.I.fits | 0.00351331 | 0.00308565 | 77.2395 | path/to/file |
# | name | rms_min | datetime | skyreg__centre_ra | skyreg__centre_dec | skyreg__xtr_radius |
# |:------------------------------|----------:|:---------------------------------|--------------------:|---------------------:|---------------------:|
# | VAST_2118-06A.EPOCH01.I.fits | 0.173946 | 2019-08-27 18:12:16.700000+00:00 | 319.652 | -6.2989 | 6.7401 |
# | VAST_2118-06A.EPOCH03x.I.fits | 0.165395 | 2019-10-29 10:01:20.500000+00:00 | 319.652 | -6.2989 | 6.7401 |
# | VAST_2118-06A.EPOCH02.I.fits | 0.16323 | 2019-10-30 08:31:20.200000+00:00 | 319.652 | -6.2989 | 6.7401 |
# Explode out the img_diff column.
extr_df = extr_df.explode('img_diff').reset_index()
total_to_extract = extr_df.shape[0]
if add_mode:
# If we are adding images to the run we assume that monitoring was
# also performed before (enforced by the pre-run checks) so now we
# only want to force extract in three situations:
# 1. Any force extraction in a new image.
# 2. The forced extraction is attached to a new source from the new
# images.
# 3. A new relation has been created and they need the forced
# measuremnts filled in (actually covered by 2.)
extr_df = (
extr_df[~extr_df['img_diff'].isin(done_images_df['name'])].append(
extr_df[(~extr_df['source'].isin(done_source_ids))
& (extr_df['img_diff'].isin(done_images_df.name))]).
sort_index())
logger.info(f"{extr_df.shape[0]} new measurements to force extract"
f" (from {total_to_extract} total)")
timer.reset()
extr_df = parallel_extraction(extr_df, images_df,
sources_df[['source', 'image', 'flux_peak']],
min_sigma, edge_buffer, cluster_threshold,
allow_nan, add_mode, p_run.path)
logger.info('Force extraction step time: %.2f seconds', timer.reset())
# make measurement names unique for db constraint
extr_df['name'] = extr_df['name'] + f'_f_run{p_run.id:06d}'
# select sensible flux values and set the columns with fix values
values = {'flux_int': 0, 'flux_int_err': 0}
extr_df = extr_df.fillna(value=values)
extr_df = extr_df[(extr_df['flux_int'] != 0)
& (extr_df['flux_int_err'] != 0)
& (extr_df['chi_squared_fit'] != np.inf)
& (extr_df['chi_squared_fit'] != np.nan)]
default_pos_err = settings.POS_DEFAULT_MIN_ERROR / 3600.
extr_df['ra_err'] = default_pos_err
extr_df['dec_err'] = default_pos_err
extr_df['err_bmaj'] = 0.
extr_df['err_bmin'] = 0.
extr_df['err_pa'] = 0.
extr_df['ew_sys_err'] = cfg_err_ra
extr_df['ns_sys_err'] = cfg_err_dec
extr_df['error_radius'] = 0.
extr_df['uncertainty_ew'] = np.hypot(cfg_err_ra, default_pos_err)
extr_df['weight_ew'] = 1. / extr_df['uncertainty_ew'].values**2
extr_df['uncertainty_ns'] = np.hypot(cfg_err_dec, default_pos_err)
extr_df['weight_ns'] = 1. / extr_df['uncertainty_ns'].values**2
extr_df['flux_peak'] = extr_df['flux_int']
extr_df['flux_peak_err'] = extr_df['flux_int_err']
extr_df['local_rms'] = extr_df['flux_int_err']
extr_df['snr'] = (extr_df['flux_peak'].values /
extr_df['local_rms'].values)
extr_df['spectral_index'] = 0.
extr_df['dr'] = 0.
extr_df['d2d'] = 0.
extr_df['forced'] = True
extr_df['compactness'] = 1.
extr_df['psf_bmaj'] = extr_df['bmaj']
extr_df['psf_bmin'] = extr_df['bmin']
extr_df['psf_pa'] = extr_df['pa']
extr_df['flag_c4'] = False
extr_df['spectral_index_from_TT'] = False
extr_df['has_siblings'] = False
extr_df['flux_int_isl_ratio'] = 1.0
extr_df['flux_peak_isl_ratio'] = 1.0
col_order = read_schema(images_df.iloc[0]['measurements_path']).names
col_order.remove('id')
remaining = list(set(extr_df.columns) - set(col_order))
extr_df = extr_df[col_order + remaining]
# upload the measurements, a column 'id' is returned with the DB id
extr_df = make_upload_measurements(extr_df)
extr_df = extr_df.rename(columns={'source_tmp_id': 'source'})
# write forced measurements to specific parquet
logger.info('Saving forced measurements to specific parquet file...')
parallel_write_parquet(extr_df, p_run.path, add_mode)
# Required to rename this column for the image add mode.
extr_df = extr_df.rename(columns={'time': 'datetime'})
# append new meas into main df and proceed with source groupby etc
sources_df = sources_df.append(
extr_df.loc[:, extr_df.columns.isin(sources_df.columns)],
ignore_index=True)
# get the number of forced extractions for the run
forced_parquets = glob(
os.path.join(p_run.path, "forced_measurements*.parquet"))
if forced_parquets:
n_forced = (dd.read_parquet(forced_parquets,
columns=['id'
]).count().compute().values[0])
else:
n_forced = 0
logger.info('Total forced extraction time: %.2f seconds',
timer.reset_init())
return sources_df, n_forced
def load_metadata(self):
schema = pq.read_schema(self.filename)
metadata_json = schema.metadata[self.custom_meta_key.encode()]
metadata = json.loads(metadata_json)
return metadata
import pyarrow as pa
import pyarrow.csv as pv
import pyarrow.parquet as pq
from pyarrow.lib import Table
csv = 'accumulated_data_300_million_rows_id_filter.csv'
target_file = '../data/accumulated_data_300_million_rows_id_filter_1mill.parquet'
csv_read_options = pv.ReadOptions(skip_rows=0,
encoding="utf8",
column_names=["unit_id"])
# Types: https://arrow.apache.org/docs/python/api/datatypes.html
data_schema = pa.schema([('unit_id', pa.uint64())])
# ConvertOptions: https://arrow.apache.org/docs/python/generated/pyarrow.csv.ConvertOptions.html#pyarrow.csv.ConvertOptions
csv_convert_options = pv.ConvertOptions(column_types=data_schema)
table: Table = pv.read_csv(input_file=csv,
read_options=csv_read_options,
convert_options=csv_convert_options)
pq.write_table(table, target_file)
print('Generated file with the following:')
print('Parquet metadata: ' + str(pq.read_metadata(target_file)))
print('Parquet schema: ' + pq.read_schema(target_file).to_string())
import pandas as pd from pyarrow.parquet import read_schema ITEM_FEATURES = True SHOP_FEATURES = True ITEM_CATEGORY_FEATURES = True CALENDAR_FEATURES = True LAGGED_FEATURES = True ROLLING_FEATURES = True MISSINGNESS_FEATURES = True EMBEDDING_FEATURES = True MEDIAN_FEATURES = True MEAN_FEATURES = True BY_SHOP_ID = True BY_ITEM_ID = True BY_ITEM_CATEGORY_ID = True BY_SHOP_ID_ITEM_ID = True BY_SHOP_ID_ITEM_CATEGORY_ID = True DATA_FILE = "../input/all_features.parquet" FEATURES = pd.Index(read_schema(DATA_FILE).names).drop(["__index_level_0__", "item_cnt_month"], errors="ignore") SELECTED_FEATURES = None EARLY_STOPPING_ROUNDS = 20
def get_cols(fs, filename):
with fs.open(filename, 'rb') as f:
schema = parquet.read_schema(f)
return set(schema.names)
def print_statistics(file):
print('Parquet metadata: ' + str(pq.read_metadata(file)))
print('Parquet schema: ' + pq.read_schema(file).to_string())
print('Size of output file on disk: ' + str(get_file_size_in_mb(file)) +
' MB')
