I have a dataframe df
loaded from Hive table and it has a timestamp column, say ts
, with string type of format dd-MMM-yy hh.mm.ss.MS a
(converted to python datetime library, this is %d-%b-%y %I.%M.%S.%f %p
).
Now I want to filter rows from the dataframe that are from the last five minutes:
only_last_5_minutes = df.filter( datetime.strptime(df.ts, '%d-%b-%y %I.%M.%S.%f %p') > datetime.now() - timedelta(minutes=5) )
However, this does not work and I get this message
TypeError: strptime() argument 1 must be string, not Column
It looks like I have wrong application of column operation and it seems to me I have to create a lambda function to filter each column that satisfies the desired condition, but being a newbie to Python and lambda expression in particular, I don't know how to create my filter correct. Please advise.
P.S. I prefer to express my filters as Python native (or SparkSQL) rather than a filter inside Hive sql query expression 'WHERE'.
preferred:
df = sqlContext.sql("SELECT * FROM my_table") df.filter( // filter here)
not preferred:
df = sqlContext.sql("SELECT * FROM my_table WHERE...")
It is possible to use user defined function.
from datetime import datetime, timedelta from pyspark.sql.types import BooleanType, TimestampType from pyspark.sql.functions import udf, col def in_last_5_minutes(now): def _in_last_5_minutes(then): then_parsed = datetime.strptime(then, '%d-%b-%y %I.%M.%S.%f %p') return then_parsed > now - timedelta(minutes=5) return udf(_in_last_5_minutes, BooleanType())
Using some dummy data:
df = sqlContext.createDataFrame([ (1, '14-Jul-15 11.34.29.000000 AM'), (2, '14-Jul-15 11.34.27.000000 AM'), (3, '14-Jul-15 11.32.11.000000 AM'), (4, '14-Jul-15 11.29.00.000000 AM'), (5, '14-Jul-15 11.28.29.000000 AM') ], ('id', 'datetime')) now = datetime(2015, 7, 14, 11, 35) df.where(in_last_5_minutes(now)(col("datetime"))).show()
And as expected we get only 3 entries:
+--+--------------------+ |id| datetime| +--+--------------------+ | 1|14-Jul-15 11.34.2...| | 2|14-Jul-15 11.34.2...| | 3|14-Jul-15 11.32.1...| +--+--------------------+
Parsing datetime string all over again is rather inefficient so you may consider storing TimestampType
instead.
def parse_dt(): def _parse(dt): return datetime.strptime(dt, '%d-%b-%y %I.%M.%S.%f %p') return udf(_parse, TimestampType()) df_with_timestamp = df.withColumn("timestamp", parse_dt()(df.datetime)) def in_last_5_minutes(now): def _in_last_5_minutes(then): return then > now - timedelta(minutes=5) return udf(_in_last_5_minutes, BooleanType()) df_with_timestamp.where(in_last_5_minutes(now)(col("timestamp")))
and result:
+--+--------------------+--------------------+ |id| datetime| timestamp| +--+--------------------+--------------------+ | 1|14-Jul-15 11.34.2...|2015-07-14 11:34:...| | 2|14-Jul-15 11.34.2...|2015-07-14 11:34:...| | 3|14-Jul-15 11.32.1...|2015-07-14 11:32:...| +--+--------------------+--------------------+
Finally it is possible to use raw SQL query with timestamps:
query = """SELECT * FROM df WHERE unix_timestamp(datetime, 'dd-MMM-yy HH.mm.ss.SSSSSS a') > {0} """.format(time.mktime((now - timedelta(minutes=5)).timetuple())) sqlContext.sql(query)
Same as above it would be more efficient to parse date strings once.
If column is already a timestamp
it possible to use datetime
literals:
from pyspark.sql.functions import lit df_with_timestamp.where( df_with_timestamp.timestamp > lit(now - timedelta(minutes=5)))
EDIT
Since Spark 1.5 you can parse date string as follows:
from pyspark.sql.functions import from_unixtime, unix_timestamp from pyspark.sql.types import TimestampType df.select((from_unixtime(unix_timestamp( df.datetime, "yy-MMM-dd h.mm.ss.SSSSSS aa" ))).cast(TimestampType()).alias("datetime"))