PySpark generating consecutive increasing index for each window - pyspark

I would like to generate consecutively increasing index ids for each dataframe window, and the index start point can be customed, say 212 for the following example.
INPUT:
+---+-------------+
| id| component|
+---+-------------+
| a|1047972020224|
| b|1047972020224|
| c|1047972020224|
| d| 670014898176|
| e| 670014898176|
| f| 146028888064|
| g| 146028888064|
+---+-------------+
EXPECTED OUTPUT:
+---+-------------+-----------------------------+
| id| component| partition_index|
+---+-------------+-----------------------------+
| a|1047972020224| 212|
| b|1047972020224| 212|
| c|1047972020224| 212|
| d| 670014898176| 213|
| e| 670014898176| 213|
| f| 146028888064| 214|
| g| 146028888064| 214|
+---+-------------+-----------------------------+
Not sure if Window.partitionBy('component').rowsBetween(Window.unboundedPreceding, Window.unboundedFollowing) can be helpful in this problem. Any ideas?


You don't have any obvious partitioning here, so you can use dense_rank with an unpartitioned window and add 211 to the result. e.g.
from pyspark.sql import functions as F, Window
df2 = df.withColumn(
'index',
F.dense_rank().over(Window.orderBy(F.desc('component'))) + 211
)
df2.show()
+---+-------------+-----+
| id| component|index|
+---+-------------+-----+
| a|1047972020224| 212|
| b|1047972020224| 212|
| c|1047972020224| 212|
| d| 670014898176| 213|
| e| 670014898176| 213|
| f| 146028888064| 214|
| g| 146028888064| 214|
+---+-------------+-----+

Related

Need to add order on ids on the basis of timestamp

Desired Outcome
Tried Everthing group by and condition but not working

You can achieve that with windowing function, like this:
from pyspark.sql.window import Window
from pyspark.sql.functions import row_number
window = Window.partitionBy("user").orderBy("timestamp")
df.withColumn("order", row_number().over(window)).show()
+----+---------+-----+
|user|timestamp|order|
+----+---------+-----+
| 111| 12:00| 1|
| 111| 12:30| 2|
| 111| 12:45| 3|
| 112| 12:00| 1|
| 112| 12:30| 2|
| 112| 12:45| 3|
| 113| 12:00| 1|
| 113| 12:30| 2|
| 113| 12:45| 3|
+----+---------+-----+

PySpark - Getting the latest date less than another given date

I need some help. I have two dataframes, one has a few dates and the other has my significant data, catalogued by date.
It goes something like this:
First df, with the relevant data
+------+----------+---------------+
| id| test_date| score|
+------+----------+---------------+
| 1|2021-03-31| 94|
| 1|2021-01-31| 93|
| 1|2020-12-31| 100|
| 1|2020-06-30| 95|
| 1|2019-10-31| 58|
| 1|2017-10-31| 78|
| 2|2020-01-31| 79|
| 2|2018-03-31| 66|
| 2|2016-05-31| 77|
| 3|2021-05-31| 97|
| 3|2020-07-31| 100|
| 3|2019-07-31| 99|
| 3|2019-06-30| 98|
| 3|2018-07-31| 91|
| 3|2018-02-28| 86|
| 3|2017-11-30| 82|
+------+----------+---------------+
Second df, with the dates
+--------------+--------------+--------------+
| eval_date_1| eval_date_2| eval_date_3|
+--------------+--------------+--------------+
| 2021-01-31| 2020-10-31| 2019-06-30|
+--------------+--------------+--------------+
Needed DF
+------+--------------+---------+--------------+---------+--------------+---------+
| id| eval_date_1| score_1 | eval_date_2| score_2 | eval_date_3| score_3 |
+------+--------------+---------+--------------+---------+--------------+---------+
| 1| 2021-01-31| 93| 2020-10-31| 95| 2019-06-30| 78|
| 2| 2021-01-31| 79| 2020-10-31| 79| 2019-06-30| 66|
| 3| 2021-01-31| 100| 2020-10-31| 100| 2019-06-30| 98|
+------+--------------+---------+--------------+---------+--------------+---------+
So, for instance, for the first id, the needed df takes the scores from the second, fourth and sixth rows from the first df. Those are the most updated dates that stay equal to or below the eval_date on the second df.

Assuming df is your main dataframe and df_date is the one which contains only dates.
from functools import reduce
from pyspark.sql import functions as F, Window as W
df_final = reduce(
lambda a, b: a.join(b, on="id"),
(
df.join(
F.broadcast(df_date.select(f"eval_date_{i}")),
on=F.col(f"eval_date_{i}") >= F.col("test_date"),
)
.withColumn(
"rnk",
F.row_number().over(W.partitionBy("id").orderBy(F.col("test_date").desc())),
)
.where("rnk=1")
.select("id", f"eval_date_{i}", "score")
for i in range(1, 4)
),
)
df_final.show()
+---+-----------+-----+-----------+-----+-----------+-----+
| id|eval_date_1|score|eval_date_2|score|eval_date_3|score|
+---+-----------+-----+-----------+-----+-----------+-----+
| 1| 2021-01-31| 93| 2020-10-31| 95| 2019-06-30| 78|
| 3| 2021-01-31| 100| 2020-10-31| 100| 2019-06-30| 98|
| 2| 2021-01-31| 79| 2020-10-31| 79| 2019-06-30| 66|
+---+-----------+-----+-----------+-----+-----------+-----+

Spark - Divide a dataframe into n number of records

I have a dataframe with 2 or more columns and 1000 records. I want to split the data into 100 records chunks randomly without any conditions.
So expected output in records count should be something like this,
[(1,2....100),(101,102,103...200),.....,(900,901...1000)]
Here's the solution that worked for my use case after trying different approaches:
https://stackoverflow.com/a/61276734/12322995

As #Shaido said randomsplit is ther for splitting dataframe is popular approach..
Thought differently about repartitionByRange with => spark 2.3
repartitionByRange public Dataset repartitionByRange(int
numPartitions,
scala.collection.Seq partitionExprs) Returns a new Dataset partitioned by the given
partitioning expressions into numPartitions. The resulting Dataset is
range partitioned. At least one partition-by expression must be
specified. When no explicit sort order is specified, "ascending nulls
first" is assumed. Parameters: numPartitions - (undocumented)
partitionExprs - (undocumented) Returns: (undocumented) Since:
2.3.0
package examples
import org.apache.log4j.Level
import org.apache.spark.sql.{Dataset, SparkSession}
object RepartitionByRange extends App {
val logger = org.apache.log4j.Logger.getLogger("org")
logger.setLevel(Level.WARN)
val spark = SparkSession.builder().appName(getClass.getName).master("local").getOrCreate()
val sc = spark.sparkContext
import spark.implicits._
val t1 = sc.parallelize(0 until 1000).toDF("id")
val repartitionedOrders: Dataset[String] = t1.repartitionByRange(10, $"id")
.mapPartitions(rows => {
val idsInPartition = rows.map(row => row.getAs[Int]("id")).toSeq.sorted.mkString(",")
Iterator(idsInPartition)
})
repartitionedOrders.show(false)
println("number of chunks or partitions :" + repartitionedOrders.rdd.getNumPartitions)
}
Result :
+---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+
|value |
+---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+
|0,1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16,17,18,19,20,21,22,23,24,25,26,27,28,29,30,31,32,33,34,35,36,37,38,39,40,41,42,43,44,45,46,47,48,49,50,51,52,53,54,55,56,57,58,59,60,61,62,63,64,65,66,67,68,69,70,71,72,73,74,75,76,77,78,79,80,81,82,83,84,85,86,87,88,89,90,91,92,93,94,95,96,97,98,99 |
|100,101,102,103,104,105,106,107,108,109,110,111,112,113,114,115,116,117,118,119,120,121,122,123,124,125,126,127,128,129,130,131,132,133,134,135,136,137,138,139,140,141,142,143,144,145,146,147,148,149,150,151,152,153,154,155,156,157,158,159,160,161,162,163,164,165,166,167,168,169,170,171,172,173,174,175,176,177,178,179,180,181,182,183,184,185,186,187,188,189,190,191,192,193,194,195,196,197,198,199|
|200,201,202,203,204,205,206,207,208,209,210,211,212,213,214,215,216,217,218,219,220,221,222,223,224,225,226,227,228,229,230,231,232,233,234,235,236,237,238,239,240,241,242,243,244,245,246,247,248,249,250,251,252,253,254,255,256,257,258,259,260,261,262,263,264,265,266,267,268,269,270,271,272,273,274,275,276,277,278,279,280,281,282,283,284,285,286,287,288,289,290,291,292,293,294,295,296,297,298,299|
|300,301,302,303,304,305,306,307,308,309,310,311,312,313,314,315,316,317,318,319,320,321,322,323,324,325,326,327,328,329,330,331,332,333,334,335,336,337,338,339,340,341,342,343,344,345,346,347,348,349,350,351,352,353,354,355,356,357,358,359,360,361,362,363,364,365,366,367,368,369,370,371,372,373,374,375,376,377,378,379,380,381,382,383,384,385,386,387,388,389,390,391,392,393,394,395,396,397,398,399|
|400,401,402,403,404,405,406,407,408,409,410,411,412,413,414,415,416,417,418,419,420,421,422,423,424,425,426,427,428,429,430,431,432,433,434,435,436,437,438,439,440,441,442,443,444,445,446,447,448,449,450,451,452,453,454,455,456,457,458,459,460,461,462,463,464,465,466,467,468,469,470,471,472,473,474,475,476,477,478,479,480,481,482,483,484,485,486,487,488,489,490,491,492,493,494,495,496,497,498,499|
|500,501,502,503,504,505,506,507,508,509,510,511,512,513,514,515,516,517,518,519,520,521,522,523,524,525,526,527,528,529,530,531,532,533,534,535,536,537,538,539,540,541,542,543,544,545,546,547,548,549,550,551,552,553,554,555,556,557,558,559,560,561,562,563,564,565,566,567,568,569,570,571,572,573,574,575,576,577,578,579,580,581,582,583,584,585,586,587,588,589,590,591,592,593,594,595,596,597,598,599|
|600,601,602,603,604,605,606,607,608,609,610,611,612,613,614,615,616,617,618,619,620,621,622,623,624,625,626,627,628,629,630,631,632,633,634,635,636,637,638,639,640,641,642,643,644,645,646,647,648,649,650,651,652,653,654,655,656,657,658,659,660,661,662,663,664,665,666,667,668,669,670,671,672,673,674,675,676,677,678,679,680,681,682,683,684,685,686,687,688,689,690,691,692,693,694,695,696,697,698,699|
|700,701,702,703,704,705,706,707,708,709,710,711,712,713,714,715,716,717,718,719,720,721,722,723,724,725,726,727,728,729,730,731,732,733,734,735,736,737,738,739,740,741,742,743,744,745,746,747,748,749,750,751,752,753,754,755,756,757,758,759,760,761,762,763,764,765,766,767,768,769,770,771,772,773,774,775,776,777,778,779,780,781,782,783,784,785,786,787,788,789,790,791,792,793,794,795,796,797,798,799|
|800,801,802,803,804,805,806,807,808,809,810,811,812,813,814,815,816,817,818,819,820,821,822,823,824,825,826,827,828,829,830,831,832,833,834,835,836,837,838,839,840,841,842,843,844,845,846,847,848,849,850,851,852,853,854,855,856,857,858,859,860,861,862,863,864,865,866,867,868,869,870,871,872,873,874,875,876,877,878,879,880,881,882,883,884,885,886,887,888,889,890,891,892,893,894,895,896,897,898,899|
|900,901,902,903,904,905,906,907,908,909,910,911,912,913,914,915,916,917,918,919,920,921,922,923,924,925,926,927,928,929,930,931,932,933,934,935,936,937,938,939,940,941,942,943,944,945,946,947,948,949,950,951,952,953,954,955,956,957,958,959,960,961,962,963,964,965,966,967,968,969,970,971,972,973,974,975,976,977,978,979,980,981,982,983,984,985,986,987,988,989,990,991,992,993,994,995,996,997,998,999|
+---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+
number of chunks or partitions : 10
UPDATE : randomsplit example :
import spark.implicits._
val t1 = sc.parallelize(0 until 1000).toDF("id")
println("With Random Split ")
val dfarray = t1.randomSplit(Array(1, 1, 1, 1, 1, 1, 1, 1, 1, 1));
println("number of dataframes " + dfarray.length + "element order is not guaranteed ")
dfarray.foreach {
df => df.show
}
Result : Will be split in to 10 dataframes and order is not guaranteed.
With Random Split
number of dataframes 10element order is not guaranteed
+---+
| id|
+---+
| 2|
| 10|
| 16|
| 30|
| 36|
| 46|
| 51|
| 91|
|100|
|121|
|136|
|138|
|149|
|152|
|159|
|169|
|198|
|199|
|220|
|248|
+---+
only showing top 20 rows
+---+
| id|
+---+
| 26|
| 40|
| 45|
| 54|
| 63|
| 72|
| 76|
|107|
|129|
|137|
|142|
|145|
|153|
|162|
|173|
|179|
|196|
|208|
|214|
|232|
+---+
only showing top 20 rows
+---+
| id|
+---+
| 7|
| 12|
| 31|
| 32|
| 38|
| 42|
| 53|
| 61|
| 68|
| 73|
| 80|
| 89|
| 96|
|115|
|117|
|118|
|131|
|132|
|139|
|146|
+---+
only showing top 20 rows
+---+
| id|
+---+
| 0|
| 24|
| 35|
| 57|
| 58|
| 65|
| 77|
| 78|
| 84|
| 86|
| 90|
| 97|
|156|
|158|
|168|
|174|
|182|
|197|
|218|
|242|
+---+
only showing top 20 rows
+---+
| id|
+---+
| 1|
| 3|
| 17|
| 18|
| 19|
| 33|
| 70|
| 71|
| 74|
| 83|
|102|
|104|
|108|
|109|
|122|
|128|
|143|
|150|
|154|
|157|
+---+
only showing top 20 rows
+---+
| id|
+---+
| 14|
| 15|
| 29|
| 44|
| 64|
| 75|
| 88|
|103|
|110|
|113|
|116|
|120|
|124|
|135|
|155|
|213|
|221|
|238|
|241|
|251|
+---+
only showing top 20 rows
+---+
| id|
+---+
| 5|
| 9|
| 21|
| 22|
| 23|
| 25|
| 27|
| 47|
| 52|
| 55|
| 60|
| 62|
| 69|
| 93|
|111|
|114|
|141|
|144|
|161|
|164|
+---+
only showing top 20 rows
+---+
| id|
+---+
| 13|
| 20|
| 39|
| 41|
| 49|
| 56|
| 67|
| 85|
| 87|
| 92|
|105|
|106|
|126|
|127|
|160|
|165|
|166|
|171|
|175|
|184|
+---+
only showing top 20 rows
+---+
| id|
+---+
| 4|
| 34|
| 50|
| 79|
| 81|
|101|
|119|
|123|
|133|
|147|
|163|
|170|
|180|
|181|
|193|
|202|
|207|
|222|
|226|
|233|
+---+
only showing top 20 rows
+---+
| id|
+---+
| 6|
| 8|
| 11|
| 28|
| 37|
| 43|
| 48|
| 59|
| 66|
| 82|
| 94|
| 95|
| 98|
| 99|
|112|
|125|
|130|
|134|
|140|
|183|
+---+
only showing top 20 rows

Since I want the data to be evenly distributed and to be able to use the chunks separately or in iterative manner using randomSplit doesn't work as it may leave empty dataframes or unequal distribution.
So using grouped can be one of the most feasible solutions here if you don't mind calling collect on your dataframe.
Eg: val newdf = df.collect.grouped(10)
That gives an Iterator[List[org.apache.spark.sql.Row]] = non-empty iterator. Can also convert it into list by adding .toList at the end
Another possible solution if we don't want Array chunks of data from the dataframe but still want to partition the data with equal counts of records we can try to use countApprox by adjusting timeout and confidence as required. Then divide that with number of records we need in a partition, which can be later used as number of partitions when using repartition or Coalesce.
countApprox instead of count because it is less expensive operation and you can feel the difference when the data size is huge
val approxCount = df.rdd.countApprox(timeout = 1000L,confidence = 0.95).getFinalValue().high
val numOfPartitions = Math.max(Math.round(approxCount / 100), 1).toInt
df.repartition(numOfPartitions)

How to find the max length unique rows from a dataframe with spark?

I am trying to find the unique rows (based on id) that have the maximum length value in a Spark dataframe. Each Column has a value of string type.
The dataframe is like:
+-----+---+----+---+---+
|id | A | B | C | D |
+-----+---+----+---+---+
|1 |toto|tata|titi| |
|1 |toto|tata|titi|tutu|
|2 |bla |blo | | |
|3 |b | c | | d |
|3 |b | c | a | d |
+-----+---+----+---+---+
The expectation is:
+-----+---+----+---+---+
|id | A | B | C | D |
+-----+---+----+---+---+
|1 |toto|tata|titi|tutu|
|2 |bla |blo | | |
|3 |b | c | a | d |
+-----+---+----+---+---+
I can't figure how to do this using Spark easily...
Thanks in advance

Note: This approach takes care of any addition/deletion of columns to the DataFrame, without the need of code change.
It can be done by first finding length of all columns after concatenating (except the first column), then filter all other rows except the row with the maximum length.
import org.apache.spark.sql.expressions._
import org.apache.spark.sql.functions._
val output = input.withColumn("rowLength", length(concat(input.columns.toList.drop(1).map(col): _*)))
.withColumn("maxLength", max($"rowLength").over(Window.partitionBy($"id")))
.filter($"rowLength" === $"maxLength")
.drop("rowLength", "maxLength")

scala> df.show
+---+----+----+----+----+
| id| A| B| C| D|
+---+----+----+----+----+
| 1|toto|tata|titi| |
| 1|toto|tata|titi|tutu|
| 2| bla| blo| | |
| 3| b| c| | d|
| 3| b| c| a| d|
+---+----+----+----+----+
scala> df.groupBy("id").agg(concat_ws("",collect_set(col("A"))).alias("A"),concat_ws("",collect_set(col("B"))).alias("B"),concat_ws("",collect_set(col("C"))).alias("C"),concat_ws("",collect_set(col("D"))).alias("D")).show
+---+----+----+----+----+
| id| A| B| C| D|
+---+----+----+----+----+
| 1|toto|tata|titi|tutu|
| 2| bla| blo| | |
| 3| b| c| a| d|
+---+----+----+----+----+

Forward-fill missing data in PySpark not working

I have a simple dataset as shown under.
| id| name| country| languages|
|1 | Bob| USA| Spanish|
|2 | Angelina| France| null|
|3 | Carl| Brazil| null|
|4 | John| Australia| English|
|5 | Anne| Nepal| null|
I am trying to impute the null values in languages with the last non-null value using pyspark.sql.window to create a window over certain rows but nothing is happening. The column which is supposed to be have null values filled, temp_filled_spark, remains unchanged i.e a copy of original languages column.
from pyspark.sql import Window
from pyspark.sql.functions import last
window = Window.partitionBy('name').orderBy('country').rowsBetween(-sys.maxsize, 0)
filled_column = last(df['languages'], ignorenulls=True).over(window)
df = df.withColumn('temp_filled_spark', filled_column)
df.orderBy('name', 'country').show(100)
I expect the output column to be:
|temp_filled_spark|
| Spanish|
| Spanish|
| Spanish|
| English|
| English|
Could anybody help pointing out the mistake?

we can create window considering entire dataframe as one partition as,
from pyspark.sql import functions as F
>>> df1.show()
+---+--------+---------+---------+
| id| name| country|languages|
+---+--------+---------+---------+
| 1| Bob| USA| Spanish|
| 2|Angelina| France| null|
| 3| Carl| Brazil| null|
| 4| John|Australia| English|
| 5| Anne| Nepal| null|
+---+--------+---------+---------+
>>> w = Window.partitionBy(F.lit(1)).orderBy(F.lit(1)).rowsBetween(-sys.maxsize, 0)
>>> df1.select("*",F.last('languages',True).over(w).alias('newcol')).show()
+---+--------+---------+---------+-------+
| id| name| country|languages| newcol|
+---+--------+---------+---------+-------+
| 1| Bob| USA| Spanish|Spanish|
| 2|Angelina| France| null|Spanish|
| 3| Carl| Brazil| null|Spanish|
| 4| John|Australia| English|English|
| 5| Anne| Nepal| null|English|
+---+--------+---------+---------+-------+
Hope this helps.!