bigquery-etl/bigquery_etl/generate_incremental_table.py

132 строки
3.7 KiB
Python

"""Run a query with a series of @submission_date values."""
from argparse import ArgumentParser
from datetime import datetime, timedelta
from functools import partial
from multiprocessing import Pool
import os.path
import subprocess
import sys
def fromisoformat(string):
"""Construct a date from the output of date.isoformat()."""
return datetime.strptime(string, "%Y-%m-%d").date()
def path(string):
"""Path to a file that must exist."""
if not os.path.exists(string):
raise ValueError()
return string
parser = ArgumentParser(description=__doc__)
parser.add_argument(
"--destination_table",
required=True,
help="The destination table flag for `bq query` to which"
" '$%%Y%%m%%d' will be added",
)
parser.add_argument(
"--start",
type=fromisoformat,
help="The first date for which the query will be submitted,"
" defaults to today in utc",
)
parser.add_argument(
"--end",
type=fromisoformat,
help="The last date for which the query will be submitted,"
" defaults to today in utc",
)
parser.add_argument(
"--max_procs",
"-P",
type=int,
default=1,
help="Maximum number of queries which may be submitted at once",
)
parser.add_argument(
"query_options",
nargs="*",
help="Arguments to pass directly to `script/entrypoint query`",
)
parser.add_argument(
"query_file", type=path, help="Path to the query file for `script/entrypoint query`"
)
def _date_range(start=None, end=None, step=None):
"""Generate dates from START to END, in steps of STEP days."""
if step is None:
if end is not None and start is not None and end < start:
step = -1
else:
step = 1
if isinstance(step, int):
step = timedelta(days=step)
if end is None:
end = datetime.utcnow().date()
if start is None:
start = end - step
for offset in range((end - start) // step):
yield start + (offset * step)
def _query(day, entrypoint, destination_table, query_options, query_file):
proc = subprocess.Popen(
[
entrypoint,
"query",
"--replace",
"--dataset_id=telemetry",
"--destination_table=" + destination_table + day.strftime("$%Y%m%d"),
"--parameter=submission_date:DATE:" + day.isoformat(),
"--max_rows=0",
"--nouse_legacy_sql",
"--quiet",
]
+ query_options
+ [query_file],
stdout=subprocess.PIPE,
stderr=subprocess.PIPE,
)
prefix = day.isoformat() + ": "
sys.stderr.write(prefix + "processing\n")
for data, dest in zip(proc.communicate(), (sys.stdout, sys.stderr)):
lines = data.split(b"\n")
if len(lines[-1]) == 0:
lines = lines[:-1]
for line in lines:
dest.write(prefix)
dest.write(line.decode())
dest.write("\n")
dest.flush()
return proc.returncode
def main():
"""Run a query with a series of @submission_date values."""
args, unknown_args = parser.parse_known_args()
# Unknown args should be passed as query_options.
args.query_options.extend(unknown_args)
target = partial(
_query,
entrypoint=os.path.join(os.path.dirname(sys.argv[0]), "entrypoint"),
destination_table=args.destination_table,
query_options=args.query_options,
query_file=args.query_file,
)
if args.max_procs == 1:
for day in _date_range(args.start, args.end):
target(day)
else:
Pool(processes=args.max_procs).map(
target, _date_range(args.start, args.end), chunksize=1
)
if __name__ == "__main__":
main()