spark join with column multiple values in list - sql

I have
Dataset A: uuid, listOfLocationsIds, name
Dataset B: locationId, latitude, longitude
A.listOfLocationIds can have multiple locationIds
How can I do a join on A and B with each value in listOfLocationsIds?
So if there are two values in listOfLocationIds, I would want the join to consider each locationId in the listOfLocationIds
A.join(B, A.listOfLocationsIds[0] == B.locationId, "left")
A.join(B, A.listOfLocationsIds[1] == B.locationId, "left")

Assume dataset A is called df with this content:
+----+-----------------+-----+
|uuid|listOfLocationsId|name |
+----+-----------------+-----+
|1 |[1, 2, 3] |name1|
|2 |[1, 3] |name1|
+----+-----------------+-----+
and dataset B is called df2 with this content:
+----------+--------+---------+
|locationId|latitude|longitude|
+----------+--------+---------+
|2 |5 |7 |
+----------+--------+---------+
And we do an array_contains join:
df = df.join(df2,
array_contains(col("listOfLocationsId"), col("locationId")), "left"
)
The final result:
+----+-----------------+-----+----------+--------+---------+
|uuid|listOfLocationsId|name |locationId|latitude|longitude|
+----+-----------------+-----+----------+--------+---------+
|1 |[1, 2, 3] |name1|2 |5 |7 |
|2 |[1, 3] |name1|null |null |null |
+----+-----------------+-----+----------+--------+---------+
Good luck!

Related

Build hierarchy based on group by column - pyspark or pandas

I have dataframe
|empid|mgrid|deptid|
|-----|-----|------|
|1 |2 |1 |
|2 |3 |1 |
|5 |6 |1 |
|2 |3 |2 |
|3 |4 |2 |
|-----|-----|------|
Expected Output
|deptid|empid|hierarchy|
|------|-----|---------|
|1 |1 |[2,3] |
|1 |2 |[3] |
|1 |5 |[6] |
|2 |2 |[3,4] |
|2 |3 |4 |
|------|-----|---------|
My Problem - I want to build the hierarchy based on deptid.
I am using the below code to build, but its not based on any column. It takes all.
import pandas as pd
def walk(df, id, f, r, prev=pd.Series(dtype="int64")):
mgr = df.loc[df[f]==id,][r]
if not mgr.isna().all():
prev = walk(df, mgr.tolist()[0], f, r, prev)
return pd.concat([mgr, prev])
Trying something like
df_pandas = df_pandas[["deptid","empid","mgrid"]]
df_pandas1 = (df_pandas.groupby("deptid"))
df_pandas1.assign(parent_lineage=lambda x: x["empid"].apply(lambda e: (walk(x, e, "empid", "mgrid")
.dropna().astype("string").tolist()))
As this is a graph problem, I would use networkx to solve it:
import networkx as nx
def get_descendents(g):
G = nx.from_pandas_edgelist(g, source='empid', target='mgrid',
create_using=nx.DiGraph)
return g['empid'].map(lambda n: list(nx.descendants(G, n)))
df['hierarchy'] = df.groupby('deptid', group_keys=False).apply(get_descendents)
Output:
empid mgrid deptid hierarchy
0 1 2 1 [2, 3]
1 2 3 1 [3]
2 5 6 1 [6]
3 2 3 2 [3, 4]
4 3 4 2 [4]
Your subgraphs:
deptid 1:
deptid 2:

Creating MAPTYPE field from multiple columns - Spark SQL

I have a use case wherein multiple keys are distributed across the dataset in a JSON format, which needs to be aggregated into a consolidated resultset for further processing.
I have been able to develop a code structure that achieves it using both Python API (PySpark) & Spark SQL, but the latter involves a more composite & tardy of doing it and involves intermediate conversations which can in the future lead to errors.
Using the below snippets, is there a better way to achieve this using Spark SQL, by creating a MAP<STRING,ARRAY<STRING> using key and value?
Data Preparation
from pyspark.sql.types import *
import pandas as pd
from io import StringIO
s = StringIO("""
id|json_struct
1|{"a":["tyeqb","",""],"e":["qwrqc","",""]}
1|{"t":["sartq","",""],"r":["fsafsq","",""]}
1|{"b":["puhqiqh","",""],"e":["hjfsaj","",""]}
2|{"b":["basajhjwa","",""],"e":["asfafas","",""]}
2|{"n":["gaswq","",""],"r":["sar","",""],"l":["sar","",""],"s":["rqqrq","",""],"m":["wrqwrq","",""]}
2|{"s":["tqqwjh","",""],"t":["afs","",""],"l":["fsaafs","",""]}
""")
df = pd.read_csv(s,delimiter='|')
sparkDF = spark.createDataFrame(df)
sparkDF.registerTempTable("INPUT")
sparkDF = sparkDF.withColumn('json_struct', F.from_json(F.col('json_struct')
,schema=MapType(StringType(),ArrayType(StringType()),True)
))
sparkDF.show(truncate=False)
+---+---------------------------------------------------------------------------------------+
|id |json_struct |
+---+---------------------------------------------------------------------------------------+
|1 |{a -> [tyeqb, , ], e -> [qwrqc, , ]} |
|1 |{t -> [sartq, , ], r -> [fsafsq, , ]} |
|1 |{b -> [puhqiqh, , ], e -> [hjfsaj, , ]} |
|2 |{b -> [basajhjwa, , ], e -> [asfafas, , ]} |
|2 |{n -> [gaswq, , ], r -> [sar, , ], l -> [sar, , ], s -> [rqqrq, , ], m -> [wrqwrq, , ]}|
|2 |{s -> [tqqwjh, , ], t -> [afs, , ], l -> [fsaafs, , ]} |
+---+---------------------------------------------------------------------------------------+
Python API (PySpark) - Implementation
As you can see, the resultant key from explode is natively a STRING type and since PySpark has create_map, which is not available within Spark SQL, it can be readily used to generate the final json_struct column ensuring a single key with a varying length ARRAYTYPE<STRING> value
sparkDF.select(
F.col('id')
,F.explode(F.col('json_struct'))
).withColumn('value',F.filter(F.col('value'), lambda x: x != '')\
).withColumn('value',F.concat_ws(',', F.col('value'))\
).groupBy('id', 'key'
).agg(F.collect_set(F.col('value')).alias('value')\
).withColumn('json_struct',F.to_json(F.create_map("key","value"))
).orderBy('id'
).show(truncate=False)
+---+---+---------------+------------------------+
|id |key|value |json_struct |
+---+---+---------------+------------------------+
|1 |a |[tyeqb] |{"a":["tyeqb"]} |
|1 |e |[hjfsaj, qwrqc]|{"e":["hjfsaj","qwrqc"]}|
|1 |r |[fsafsq] |{"r":["fsafsq"]} |
|1 |b |[puhqiqh] |{"b":["puhqiqh"]} |
|1 |t |[sartq] |{"t":["sartq"]} |
|2 |b |[basajhjwa] |{"b":["basajhjwa"]} |
|2 |n |[gaswq] |{"n":["gaswq"]} |
|2 |t |[afs] |{"t":["afs"]} |
|2 |s |[tqqwjh, rqqrq]|{"s":["tqqwjh","rqqrq"]}|
|2 |e |[asfafas] |{"e":["asfafas"]} |
|2 |l |[sar, fsaafs] |{"l":["sar","fsaafs"]} |
|2 |r |[sar] |{"r":["sar"]} |
|2 |m |[wrqwrq] |{"m":["wrqwrq"]} |
+---+---+---------------+------------------------+
Spark SQL - Implementation
Within this implementation, I have to take additional steps to ensure both key and value columns are of ARRAYTYPE and consistent lengths as map_from_arrays takes in arrays as inputs.
Is there a way to bypass these and create a similar schema as depicted using Python API?
sql.sql("""
SELECT
id,
KEY,
VALUE,
TO_JSON(MAP_FROM_ARRAYS(KEY,VALUE)) as json_struct
FROM (
SELECT
id,
key,
ARRAY(COLLECT_SET( value )) as value -- <------- ### Ensuring Value is NESTED ARRAY
FROM (
SELECT
id,
SPLIT(k,'|',1) as key, -- <------- ### Ensuring Key is Array
CONCAT_WS(',',FILTER(v,x -> x != '')) as value
FROM (
SELECT
id,
EXPLODE(FROM_JSON(json_struct,'MAP<STRING,ARRAY<STRING>>')) as (k,v)
FROM INPUT
)
)
GROUP BY 1,2
)
ORDER BY 1
""").show(truncate=False)
+---+---+-----------------+------------------------+
|id |KEY|VALUE |json_struct |
+---+---+-----------------+------------------------+
|1 |[a]|[[tyeqb]] |{"a":["tyeqb"]} |
|1 |[e]|[[hjfsaj, qwrqc]]|{"e":["hjfsaj","qwrqc"]}|
|1 |[b]|[[puhqiqh]] |{"b":["puhqiqh"]} |
|1 |[r]|[[fsafsq]] |{"r":["fsafsq"]} |
|1 |[t]|[[sartq]] |{"t":["sartq"]} |
|2 |[n]|[[gaswq]] |{"n":["gaswq"]} |
|2 |[b]|[[basajhjwa]] |{"b":["basajhjwa"]} |
|2 |[t]|[[afs]] |{"t":["afs"]} |
|2 |[s]|[[tqqwjh, rqqrq]]|{"s":["tqqwjh","rqqrq"]}|
|2 |[e]|[[asfafas]] |{"e":["asfafas"]} |
|2 |[l]|[[sar, fsaafs]] |{"l":["sar","fsaafs"]} |
|2 |[r]|[[sar]] |{"r":["sar"]} |
|2 |[m]|[[wrqwrq]] |{"m":["wrqwrq"]} |
+---+---+-----------------+------------------------+
Spark SQL instead of create_map has map. Your PySpark code could be translated into this:
df = spark.sql("""
WITH
TBL2 (SELECT id, EXPLODE(FROM_JSON(json_struct,'MAP<STRING,ARRAY<STRING>>')) from INPUT),
TBL3 (SELECT id, key, FLATTEN(COLLECT_SET(FILTER(value, x -> x != ''))) value
FROM TBL2
GROUP BY id, key)
SELECT *, TO_JSON(MAP(key, value)) json_struct
FROM TBL3
""")
df.show(truncate=0)
# +---+---+---------------+------------------------+
# |id |key|value |json_struct |
# +---+---+---------------+------------------------+
# |1 |a |[tyeqb] |{"a":["tyeqb"]} |
# |1 |e |[qwrqc, hjfsaj]|{"e":["qwrqc","hjfsaj"]}|
# |1 |b |[puhqiqh] |{"b":["puhqiqh"]} |
# |1 |r |[fsafsq] |{"r":["fsafsq"]} |
# |1 |t |[sartq] |{"t":["sartq"]} |
# |2 |b |[basajhjwa] |{"b":["basajhjwa"]} |
# |2 |n |[gaswq] |{"n":["gaswq"]} |
# |2 |s |[rqqrq, tqqwjh]|{"s":["rqqrq","tqqwjh"]}|
# |2 |t |[afs] |{"t":["afs"]} |
# |2 |e |[asfafas] |{"e":["asfafas"]} |
# |2 |l |[fsaafs, sar] |{"l":["fsaafs","sar"]} |
# |2 |r |[sar] |{"r":["sar"]} |
# |2 |m |[wrqwrq] |{"m":["wrqwrq"]} |
# +---+---+---------------+------------------------+

pyspark dataframe column creation

I'm beginner in pyspark. I have this problem where I have a vector column / list of values.
col = ["True", "False", "True"]
I want to create a column in dataframe (with 3 rows) with this vector / list of values. E.g., in pandas we can do df['col_name'] = col.
Unfortunately Spark don't have such function that work well in Pandas, you can can still achieve it by using joining. Assuming you have a sorted dataframe and a list of new value that want to be the new column:
from pyspark.sql import SparkSession
from pyspark.sql.window import Window
df = spark.createDataFrame([('a', ), ('b', ), ('c', )], ['column'])
df.show(3, False)
+------+----------+
|column|row_number|
+------+----------+
|a |1 |
|b |2 |
|c |3 |
+------+----------+
You can add a row number and do the joining:
new_add_column = spark.createDataFrame([(True, ), (False, ), (True, )], ['new_create_column'])\
.withColumn('row_number', func.row_number().over(Window.orderBy(func.lit(''))))
new_add_column.show(3, False)
+------+----------+
|column|row_number|
+------+----------+
|a |1 |
|b |2 |
|c |3 |
+------+----------+
new_add_column = spark.createDataFrame([(True, ), (False, ), (True, )], ['new_create_column'])\
.withColumn('row_number', func.row_number().over(Window.orderBy(func.lit(''))))
new_add_column.show(3, False)
+-----------------+----------+
|new_create_column|row_number|
+-----------------+----------+
|true |1 |
|false |2 |
|true |3 |
+-----------------+----------+
output = df.join(new_add_column, on='row_number', how='inner')
output.show(3, False)
+----------+------+-----------------+
|row_number|column|new_create_column|
+----------+------+-----------------+
|1 |a |true |
|2 |b |false |
|3 |c |true |
+----------+------+-----------------+

In SQL, query a table by transposing column results

Background
Forgive the title of this question, as I'm not really sure how to describe what I'm trying to do.
I have a SQL table, d, that looks like this:
+--+---+------------+------------+
|id|sex|event_type_1|event_type_2|
+--+---+------------+------------+
|a |m |1 |1 |
|b |f |0 |1 |
|c |f |1 |0 |
|d |m |0 |1 |
+--+---+------------+------------+
The Problem
I'm trying to write a query that yields the following summary of counts of event_type_1 and event_type_2 cut (grouped?) by sex:
+-------------+-----+-----+
| | m | f |
+-------------+-----+-----+
|event_type_1 | 1 | 1 |
+-------------+-----+-----+
|event_type_2 | 2 | 1 |
+-------------+-----+-----+
The thing is, this seems to involve some kind of transposition of the 2 event_type columns into rows of the query result that I'm not familiar with as a novice SQL user.
What I've tried
I've so far come up with the following query:
SELECT event_type_1, event_type_2, count(sex)
FROM d
group by event_type_1, event_type_2
But that only gives me this:
+------------+------------+-----+
|event_type_1|event_type_2|count|
+------------+------------+-----+
|1 |1 |1 |
|1 |0 |1 |
|0 |1 |2 |
+------------+------------+-----+
You can use a lateral join to unpivot the data. Then use conditional aggregate to calculate m and f:
select v.which,
count(*) filter (where d.sex = 'm') as m,
count(*) filter (where d.sex = 'f') as f
from d cross join lateral
(values (d.event_type_1, 'event_type_1'),
(d.event_type_2, 'event_type_2')
) v(val, which)
where v.val = 1
group by v.which;
Here is a db<>fiddle.

How to compare two identically structured dataframes to calculate the row differences

I've the following two identically structurred dataframes with id in common.
val originalDF = Seq((1,"gaurav","jaipur",550,70000),(2,"sunil","noida",600,80000),(3,"rishi","ahmedabad",510,65000))
.toDF("id","name","city","credit_score","credit_limit")
scala> originalDF.show(false)
+---+------+---------+------------+------------+
|id |name |city |credit_score|credit_limit|
+---+------+---------+------------+------------+
|1 |gaurav|jaipur |550 |70000 |
|2 |sunil |noida |600 |80000 |
|3 |rishi |ahmedabad|510 |65000 |
+---+------+---------+------------+------------+
val changedDF= Seq((1,"gaurav","jaipur",550,70000),(2,"sunil","noida",650,90000),(4,"Joshua","cochin",612,85000))
.toDF("id","name","city","credit_score","credit_limit")
scala> changedDF.show(false)
+---+------+------+------------+------------+
|id |name |city |credit_score|credit_limit|
+---+------+------+------------+------------+
|1 |gaurav|jaipur|550 |70000 |
|2 |sunil |noida |650 |90000 |
|4 |Joshua|cochin|612 |85000 |
+---+------+------+------------+------------+
Hence I wrote one udf to calulate the change in column values.
val diff = udf((col: String, c1: String, c2: String) => if (c1 == c2) "" else col )
val somedf=changedDF.alias("a").join(originalDF.alias("b"), col("a.id") === col("b.id")).withColumn("diffcolumn", split(concat_ws(",",changedDF.columns.map(x => diff(lit(x), changedDF(x), originalDF(x))):_*),","))
scala> somedf.show(false)
+---+------+------+------------+------------+---+------+------+------------+------------+----------------------------------+
|id |name |city |credit_score|credit_limit|id |name |city |credit_score|credit_limit|diffcolumn |
+---+------+------+------------+------------+---+------+------+------------+------------+----------------------------------+
|1 |gaurav|jaipur|550 |70000 |1 |gaurav|jaipur|550 |70000 |[, , , , ] |
|2 |sunil |noida |650 |90000 |2 |sunil |noida |600 |80000 |[, , , credit_score, credit_limit]|
+---+------+------+------------+------------+---+------+------+------------+------------+----------------------------------+
But I'm not able to get id and diffcolumn separately. If I do a
somedf.select('id) it gives me ambiguity error coz there are two ids in the joined table
I want to get all the name of the columns in any array and id corresponding to which the values have changed. Like in the changedDF credit score and credit limit of id=2,name=sunil has been changed.
Hence I wanted the resultant dataframe to give me result like
+--+---+------+------+------------+------------+---+
|id | diffcolumn |
+---+------+------+------------+------------+---
|2 |[, , , credit_score, credit_limit] |
+---+------+------+------------+------------+---+
Can anyone suggest me what approach to follow to get eh id and changed column separately in a dataframe.
For your reference, these kinds of diffs can easily be done with the spark-extension package.
It provides the diff transformation that builds that complex query for you:
import uk.co.gresearch.spark.diff._
val options = DiffOptions.default.withChangeColumn("changes") // needed to get the optional 'changes' column
val diff = originalDF.diff(changedDF, options, "id")
diff.show(false)
+----+----------------------------+---+---------+----------+---------+----------+-----------------+------------------+-----------------+------------------+
|diff|changes |id |left_name|right_name|left_city|right_city|left_credit_score|right_credit_score|left_credit_limit|right_credit_limit|
+----+----------------------------+---+---------+----------+---------+----------+-----------------+------------------+-----------------+------------------+
|N |[] |1 |gaurav |gaurav |jaipur |jaipur |550 |550 |70000 |70000 |
|I |null |4 |null |Joshua |null |cochin |null |612 |null |85000 |
|C |[credit_score, credit_limit]|2 |sunil |sunil |noida |noida |600 |650 |80000 |90000 |
|D |null |3 |rishi |null |ahmedabad|null |510 |null |65000 |null |
+----+----------------------------+---+---------+----------+---------+----------+-----------------+------------------+-----------------+------------------+
diff.select($"id", $"diff", $"changes").show(false)
+---+----+----------------------------+
|id |diff|changes |
+---+----+----------------------------+
|1 |N |[] |
|4 |I |null |
|2 |C |[credit_score, credit_limit]|
|3 |D |null |
+---+----+----------------------------+
While this is a simple example, diffing DataFrames can become complicated when wide schemas and null values are involved.
That package is well-tested, so you don't have to worry about getting that query right yourself.
Try this :
val aliasedChangedDF = changedDF.as("a")
val aliasedOriginalDF = originalDF.as("b")
val diff = udf((col: String, c1: String, c2: String) => if (c1 == c2) "" else col )
val somedf=aliasedChangedDF.join(aliasedOriginalDF, col("a.id") === col("b.id")).withColumn("diffcolumn", split(concat_ws(",",changedDF.columns.map(x => diff(lit(x), changedDF(x), originalDF(x))):_*),","))
somedf.select(col("a.id").as("id"),col("diffcolumn"))
Just change your join condition from col("a.id") === col("b.id") to "id"
Then, there will be only a single id column.
Further, you don't need the alias("a") and alias("b"). So your join simplifies from
changedDF.alias("a").join(originalDF.alias("b"), col("a.id") === col("b.id"))
to
changedDF.join(originalDF, "id")