Adding column with rolling latest prior in PySpark - python

I have a pyspark dataframe with a list of customers, days, and transaction types.
+----------+-----+------+
| Customer | Day | Type |
+----------+-----+------+
| A | 2 | X11 |
| A | 4 | X2 |
| A | 9 | Y4 |
| A | 11 | X1 |
| B | 3 | Y4 |
| B | 7 | X1 |
+----------+-----+------+
I'd like to create a column that has "most recent X type" for each customer, like so:
+----------+-----+------+-------------+
| Customer | Day | Type | MostRecentX |
+----------+-----+------+-------------+
| A | 2 | X11 | X11 |
| A | 4 | X2 | X2 |
| A | 9 | Y4 | X2 |
| A | 11 | X1 | X1 |
| B | 3 | Y4 | - |
| B | 7 | X1 | X1 |
+----------+-----+------+-------------+
So for the X types it just takes the one from the current row, but for the Y type it takes the type from the most recent X row for that member (and if there isn't one, it gets a blank or something). I imagine I need a sort of window function but not very familiar with PySpark.

You can achieve by this taking the last column that startswith the letter "X" over a Window that partitions by the Customer and orders by the Day. Specify the Window to start at the beginning of the partition and stop at the current row.
from pyspark.sql import Window
from pyspark.sql.functions import col, last, when
w = Window.partitionBy("Customer").orderBy("Day").rowsBetween(Window.unboundedPreceding, 0)
df = df.withColumn(
"MostRecentX",
last(when(col("Type").startswith("X"), col("Type")), ignorenulls=True).over(w)
)
df.show()
#+--------+---+----+-----------+
#|Customer|Day|Type|MostRecentX|
#+--------+---+----+-----------+
#| A| 2| X11| X11|
#| A| 4| X2| X2|
#| A| 9| Y4| X2|
#| A| 11| X1| X1|
#| B| 3| Y4| null|
#| B| 7| X1| X1|
#+--------+---+----+-----------+
The trick here is to use when to return the Type column only if it starts with "X". By default, when will return null. Then we can use last with ignorenulls=True to get the value for MostRecentX.
If you want to replace the null with "-" as shown in your question, just call fillna on the MostRecentX column:
df.fillna("-", subset=["MostRecentX"]).show()
#+--------+---+----+-----------+
#|Customer|Day|Type|MostRecentX|
#+--------+---+----+-----------+
#| A| 2| X11| X11|
#| A| 4| X2| X2|
#| A| 9| Y4| X2|
#| A| 11| X1| X1|
#| B| 3| Y4| -|
#| B| 7| X1| X1|
#+--------+---+----+-----------+

Related

py4JJavaError. pyspark applying function to column

spark = 2.x
New to pyspark.
While encoding date related columns for training DNN keep on facing error mentioned in the title.
from df
day month ...
1 1
2 3
3 1 ...
I am trying to get cos, sine value for each column in order to capture their cyclic nature.
When applying function to column in pyspark udf worked fine until now. But below code doesn't work
def to_cos(x, _max):
return np.sin(2*np.pi*x / _max)
to_cos_udf = udf(to_cos, DecimalType())
df = df.withColumn("month", to_cos_udf("month", 12))
I've tried it with IntegerType and tried it with only one variable def to_cos(x) however none of them seem to work and outputs:
Py4JJavaError: An error occurred while calling 0.24702.showString.
Since you havent shared the entire Stacktrack from the error , not sure what is the actual error which is causing the failure
However by the code snippets you have shared , Firstly you need to update your UDF definition as below -
Will passing arguments to a UDF function using it with lambda is probably the best approach towards it , apart from that you can use partial
Data Preparation
df = pd.DataFrame({
'month':[i for i in range(0,12)],
})
sparkDF = sql.createDataFrame(df)
sparkDF.show()
+-----+
|month|
+-----+
| 0|
| 1|
| 2|
| 3|
| 4|
| 5|
| 6|
| 7|
| 8|
| 9|
| 10|
| 11|
+-----+
Custom UDF
def to_cos(x,_max):
try:
res = np.sin(2*np.pi*x / _max)
except Exception as e:
res = 0.0
return float(res)
max_cos = 12
to_cos_udf = F.udf(lambda x: to_cos(x,max_cos),FloatType())
sparkDF = sparkDF.withColumn('month_cos',to_cos_udf('month'))
sparkDF.show()
+-----+-------------+
|month| month_cos|
+-----+-------------+
| 0| 0.0|
| 1| 0.5|
| 2| 0.8660254|
| 3| 1.0|
| 4| 0.8660254|
| 5| 0.5|
| 6|1.2246469E-16|
| 7| -0.5|
| 8| -0.8660254|
| 9| -1.0|
| 10| -0.8660254|
| 11| -0.5|
+-----+-------------+
Custom UDF - Partial
from functools import partial
partial_func = partial(to_cos,_max=max_cos)
to_cos_partial_udf = F.udf(partial_func)
sparkDF = sparkDF.withColumn('month_cos',to_cos_partial_udf('month'))
sparkDF.show()
+-----+--------------------+
|month| month_cos|
+-----+--------------------+
| 0| 0.0|
| 1| 0.49999999999999994|
| 2| 0.8660254037844386|
| 3| 1.0|
| 4| 0.8660254037844388|
| 5| 0.49999999999999994|
| 6|1.224646799147353...|
| 7| -0.4999999999999998|
| 8| -0.8660254037844384|
| 9| -1.0|
| 10| -0.8660254037844386|
| 11| -0.5000000000000004|
+-----+--------------------+

Reshape pyspark dataframe to show moving window of item interactions

I have a large pyspark dataframe of subject interactions in long format--each row describes a subject interacting with some item of interest, along with a timestamp and a rank-order for that subject's interaction (i.e., first interaction is 1, second is 2, etc.). Here's a few rows:
+----------+---------+----------------------+--------------------+
| date|itemId |interaction_date_order| userId|
+----------+---------+----------------------+--------------------+
|2019-07-23| 10005880| 1|37 |
|2019-07-23| 10005903| 2|37 |
|2019-07-23| 10005903| 3|37 |
|2019-07-23| 12458442| 4|37 |
|2019-07-26| 10005903| 5|37 |
|2019-07-26| 12632813| 6|37 |
|2019-07-26| 12632813| 7|37 |
|2019-07-26| 12634497| 8|37 |
|2018-11-24| 12245677| 1|5 |
|2018-11-24| 12245677| 1|5 |
|2019-07-29| 12541871| 2|5 |
|2019-07-29| 12541871| 3|5 |
|2019-07-30| 12626854| 4|5 |
|2019-08-31| 12776880| 5|5 |
|2019-08-31| 12776880| 6|5 |
+----------+---------+----------------------+--------------------+
I need to reshape these data such that, for each subject, a row has a length-5 moving window of interactions. So then, something like this:
+------+--------+--------+--------+--------+--------+
|userId| i-2 | i-1 | i | i+1 | i+2|
+------+--------+--------+--------+--------+--------+
|37 |10005880|10005903|10005903|12458442|10005903|
|37 |10005903|10005903|12458442|10005903|12632813|
Does anyone have suggestions for how I might do this?
Import spark and everything
from pyspark.sql import *
from pyspark.context import SparkContext
from pyspark.sql.session import SparkSession
sc = SparkContext('local')
spark = SparkSession(sc)
Create your dataframe
columns = '| date|itemId |interaction_date_order| userId|'.split('|')
lines = '''2019-07-23| 10005880| 1|37 |
2019-07-23| 10005903| 2|37 |
2019-07-23| 10005903| 3|37 |
2019-07-23| 12458442| 4|37 |
2019-07-26| 10005903| 5|37 |
2019-07-26| 12632813| 6|37 |
2019-07-26| 12632813| 7|37 |
2019-07-26| 12634497| 8|37 |
2018-11-24| 12245677| 1|5 |
2018-11-24| 12245677| 2|5 |
2019-07-29| 12541871| 3|5 |
2019-07-29| 12541871| 4|5 |
2019-07-30| 12626854| 5|5 |
2019-08-31| 12776880| 6|5 |
2019-08-31| 12776880| 7|5 |'''
Interaction = Row("date", "itemId", "interaction_date_order", "userId")
interactions = []
for line in lines.split('\n'):
column_values = line.split('|')
interaction = Interaction(column_values[0], int(column_values[1]), int(column_values[2]), int(column_values[3]))
interactions.append(interaction)
df = spark.createDataFrame(interactions)
now we have
df.show()
+----------+--------+----------------------+------+
| date| itemId|interaction_date_order|userId|
+----------+--------+----------------------+------+
|2019-07-23|10005880| 1| 37|
|2019-07-23|10005903| 2| 37|
|2019-07-23|10005903| 3| 37|
|2019-07-23|12458442| 4| 37|
|2019-07-26|10005903| 5| 37|
|2019-07-26|12632813| 6| 37|
|2019-07-26|12632813| 7| 37|
|2019-07-26|12634497| 8| 37|
|2018-11-24|12245677| 1| 5|
|2018-11-24|12245677| 2| 5|
|2019-07-29|12541871| 3| 5|
|2019-07-29|12541871| 4| 5|
|2019-07-30|12626854| 5| 5|
|2019-08-31|12776880| 6| 5|
|2019-08-31|12776880| 7| 5|
+----------+--------+----------------------+------+
Create a window and collect itemId with count
from pyspark.sql.window import Window
import pyspark.sql.functions as F
window = Window() \
.partitionBy('userId') \
.orderBy('interaction_date_order') \
.rowsBetween(Window.currentRow, Window.currentRow+4)
df2 = df.withColumn("itemId_list", F.collect_list('itemId').over(window))
df2 = df2.withColumn("itemId_count", F.count('itemId').over(window))
df_final = df2.where(df2['itemId_count'] == 5)
now we have
df_final.show()
+----------+--------+----------------------+------+--------------------+------------+
| date| itemId|interaction_date_order|userId| itemId_list|itemId_count|
+----------+--------+----------------------+------+--------------------+------------+
|2018-11-24|12245677| 1| 5|[12245677, 122456...| 5|
|2018-11-24|12245677| 2| 5|[12245677, 125418...| 5|
|2019-07-29|12541871| 3| 5|[12541871, 125418...| 5|
|2019-07-23|10005880| 1| 37|[10005880, 100059...| 5|
|2019-07-23|10005903| 2| 37|[10005903, 100059...| 5|
|2019-07-23|10005903| 3| 37|[10005903, 124584...| 5|
|2019-07-23|12458442| 4| 37|[12458442, 100059...| 5|
+----------+--------+----------------------+------+--------------------+------------+
Final touch
df_final2 = (df_final
.withColumn('i-2', df_final['itemId_list'][0])
.withColumn('i-1', df_final['itemId_list'][1])
.withColumn('i', df_final['itemId_list'][2])
.withColumn('i+1', df_final['itemId_list'][3])
.withColumn('i+2', df_final['itemId_list'][4])
.select('userId', 'i-2', 'i-1', 'i', 'i+1', 'i+2')
)
df_final2.show()
+------+--------+--------+--------+--------+--------+
|userId| i-2| i-1| i| i+1| i+2|
+------+--------+--------+--------+--------+--------+
| 5|12245677|12245677|12541871|12541871|12626854|
| 5|12245677|12541871|12541871|12626854|12776880|
| 5|12541871|12541871|12626854|12776880|12776880|
| 37|10005880|10005903|10005903|12458442|10005903|
| 37|10005903|10005903|12458442|10005903|12632813|
| 37|10005903|12458442|10005903|12632813|12632813|
| 37|12458442|10005903|12632813|12632813|12634497|
+------+--------+--------+--------+--------+--------+

Find column names of interconnected row values - Spark

I have a Spark dataframe that adheres to the following structure:
+------+-----------+-----------+-----------+------+
|ID | Name1 | Name2 | Name3 | Y |
+------+-----------+-----------+-----------+------+
| 1 | A,1 | B,1 | C,4 | B |
| 2 | D,2 | E,2 | F,8 | D |
| 3 | G,5 | H,2 | I,3 | H |
+------+-----------+-----------+-----------+------+
For every row I want to find in which column the value of Y is denoted as the first element. So, ideally I want to retrieve a list like: [Name2,Name1,Name2].
I am not sure how and whether it works to convert first to a RDD, then use a map function and convert the result back to DataFrame.
Any ideas are welcome.
You can probably try this piece of code :
df.show()
+---+-----+-----+-----+---+
| ID|Name1|Name2|Name3| Y|
+---+-----+-----+-----+---+
| 1| A,1| B,1| C,4| B|
| 2| D,2| E,2| F,8| D|
| 3| G,5| H,2| I,3| H|
+---+-----+-----+-----+---+
from pyspark.sql import functions as F
name_cols = ["Name1", "Name2", "Name3"]
cond = F
for col in name_cols:
cond = cond.when(F.split(F.col(col),',').getItem(0) == F.col("Y"), col)
df.withColumn("whichName", cond).show()
+---+-----+-----+-----+---+---------+
| ID|Name1|Name2|Name3| Y|whichName|
+---+-----+-----+-----+---+---------+
| 1| A,1| B,1| C,4| B| Name2|
| 2| D,2| E,2| F,8| D| Name1|
| 3| G,5| H,2| I,3| H| Name2|
+---+-----+-----+-----+---+---------+

Pyspark | Seperate the string / int values from the dataframe

I have a Spark Dataframe as below:
+---------+
|col_str_1|
+---------+
| 1|
| 2|
| 3|
| 4|
| 5|
| 6|
| 7|
| 8|
| 9|
| a|
| b|
| c|
| d|
| e|
| f|
| g|
| h|
| 1|
| 2|
| 3.0|
+---------+
I want to separate the string / int / float values based on request
For Example:
Req is for STRING, return DF must be like below
+---------+
|col_str_1|
+---------+
| a|
| b|
| c|
| d|
| e|
| f|
| g|
| h|
+---------+
Req is for Integer, return DF must be like below
+---------+
|col_str_1|
+---------+
| 1|
| 2|
| 3|
| 4|
| 5|
| 6|
| 7|
| 8|
| 9|
| 1|
| 2|
+---------+
Tried below steps:
>> df = sqlContext.sql('select * from --db--.vt_prof_test')
>> columns = df.columns[0]
>> df.select(columns).????
how to proceed further, either use filter or map, can any one help me out ??
You can go for udf
import pyspark.sql.functions as F
df = sqlContext.sql('select * from --db--.vt_prof_test')
REQUEST = 'STRING'
request_bc = sc.broadcast(REQUEST)
def check_value(val):
if request_bc.value == 'STRING':
try:
val = int(val)
return None
except:
return val
if request_bc.value == 'INTEGER':
try:
val = int(val)
return val
except:
return None
check_udf = F.udf(lambda x: check_value(x))
df = df.select(check_udf(F.col('col_str_1').alias('col_str_1')).dropna()
Set the REQUEST parameter according to the need.

Transpose mutiple columns in a Pyspark dataframe with a condition

I have a spark dataframe that looks like this.
id cd1 version1 dt1 cd2 version2 dt2 cd3 version3 dt3
1 100 1 20100101 101 1 20100101 102 20100301
1 101 1 20100102 102 20100201 100 1 20100302
2 201 1 20100103 100 1 20100301 100 1 20100303
2 202 2 20100104 100 1 20100105
I need to transpose all the codes into a single column with the following conditions
If the corresponding version code is 1, add a decimal point after the first digit
Each patient should have distinct codes
For the above example, the output should look like this.
id code dt
1 1.00 20100101
1 1.01 20100101
1 102 20100301
1 1.01 20100102
1 102 20100201
1 10.0 20100302
2 2.01 20100103
2 1.00 20100301
2 1.00 20100303
2 202 20100104
2 10.0 20100105
I am using Pyspark to do this. In the above example, I have shown only 3 codes with their corresponding version columns but I have 30 such columns. Also, this data has around 25 million rows.
Any ideas on how to accomplish this will be extremely helpful.
You can explode a list of these columns so that there is only one (cd, version) pair per line
First, let's create the dataframe:
df = sc.parallelize([[1,100,1,101,1,102,None],[1,101,1,102,None,100,1],[2,201,1,100,1,100,1],
[2,202,2,100,1,None,None]]).toDF(["id","cd1","version1","cd2","version2","cd3","version3"])
Using posexplode:
import pyspark.sql.functions as psf
from itertools import chain
nb_versions = 4
df = df.na.fill(-1).select(
"id",
psf.posexplode(psf.create_map(list(chain(*[(psf.col("cd" + str(i)), psf.col("version"+str(i))) for i in range(1, nb_versions)])))).alias("pos", "cd", "version")
).drop("pos").filter("cd != -1")
+---+---+-------+
| id| cd|version|
+---+---+-------+
| 1|100| 1|
| 1|101| 1|
| 1|102| -1|
| 1|101| 1|
| 1|102| -1|
| 1|100| 1|
| 2|201| 1|
| 2|100| 1|
| 2|100| 1|
| 2|202| 2|
| 2|100| 1|
+---+---+-------+
Using explode:
nb_versions = 4
df = df.select(
"id",
psf.explode(psf.array(
[psf.struct(
psf.col("cd" + str(i)).alias("cd"),
psf.col("version" + str(i)).alias("version")) for i in range(1, nb_versions)])).alias("temp"))\
.select("id", "temp.*")
+---+----+-------+
| id| cd|version|
+---+----+-------+
| 1| 100| 1|
| 1| 101| 1|
| 1| 102| null|
| 1| 101| 1|
| 1| 102| null|
| 1| 100| 1|
| 2| 201| 1|
| 2| 100| 1|
| 2| 100| 1|
| 2| 202| 2|
| 2| 100| 1|
| 2|null| null|
+---+----+-------+
Now we can implement your conditions
division by 100 for version==1
distinct values
We'll use functions when, otherwise for the condition and distinct:
df.withColumn("cd", psf.when(df.version == 1, df.cd/100).otherwise(df.cd))\
.distinct().drop("version")
+---+-----+
| id| cd|
+---+-----+
| 1| 1.0|
| 1| 1.01|
| 1|102.0|
| 2| 1.0|
| 2| 2.01|
| 2|202.0|
+---+-----+
This is how I did it. I am sure there is a better way to do this.
def process_code(raw_data):
for i in range(1,4):
cd_col_name = "cd" + str(i)
version_col_name = "version" + str(i)
raw_data = raw_data.withColumn("mod_cd" + str(i), when(raw_data[version_col_name] == 1, concat(substring(raw_data[cd_col_name],1,1),lit("."),substring(raw_data[cd_col_name],2,20))).otherwise(raw_data[cd_col_name]))
mod_cols = [col for col in raw_data.columns if 'mod_cd' in col]
nb_versions = 3
new = raw_data.fillna('9999', subset=mod_cols).select("id", psf.posexplode(psf.create_map(list(chain(*[(psf.col("mod_cd" + str(i)), psf.col("dt"+str(i))) for i in range(1, nb_versions)])))).alias("pos", "final_cd", "final_date")).drop("pos")
return new
test = process_code(df)
test = test.filter(test.final_cd != '9999')
test.show(100, False)

Categories

Resources