Topics Covered
Spark dataframe add row number is very common requirement especially if you are working on ELT in Spark. You can use monotonically_increasing_id method to generate incremental numbers. However the numbers won't be consecutive if the dataframe has more than 1 partition. Let's see a simple example to understand it :
scala> df_usa.show(false) +-----------+----------+-------------+ |state_name |state_abbr|state_capital| +-----------+----------+-------------+ |Alabama |AL |Montgomery | |Alaska |AK |Juneau | |Arizona |AZ |Phoenix | |Arkansas |AR |Little Rock | |California |CA |Sacramento | |Colorado |CO |Denver | |Connecticut|CT |Hartford | |Delaware |DE |Dover | |Florida |FL |Tallahassee | |Georgia |GA |Atlanta | |Hawaii |HI |Honolulu | |Idaho |ID |Boise | |Illinois |IL |Springfield | |Indiana |IN |Indianapolis | |Iowa |IA |Des Moines | |Kansas |KS |Topeka | |Kentucky |KY |Frankfort | |Louisiana |LA |Baton Rouge | |Maine |ME |Augusta | |Maryland |MD |Annapolis | +-----------+----------+-------------+ only showing top 20 rows
So I have a dataframe which has information about all 50 states in USA. Now I want to add a new column "state_id" as sequence number. Before that I will create 3 version of this dataframe with 1,2 & 3 partitions respectively. This is required to understand behavior of monotonically_increasing_id clearly.
scala> val df_usa1 = df_usa.repartition(1) df_usa1: org.apache.spark.sql.Dataset[org.apache.spark.sql.Row] = [state_name: string, state_abbr: string ... 1 more field] scala> val df_usa2 = df_usa.repartition(2) df_usa2: org.apache.spark.sql.Dataset[org.apache.spark.sql.Row] = [state_name: string, state_abbr: string ... 1 more field] scala> val df_usa3 = df_usa.repartition(3) df_usa3: org.apache.spark.sql.Dataset[org.apache.spark.sql.Row] = [state_name: string, state_abbr: string ... 1 more field]
Now let's generate the numbers and compare the output.
monotonically_increasing_id – single partition dataframe
scala> df_usa1.withColumn("state_id",monotonically_increasing_id).show(50)
+--------------+----------+--------------+--------+
| state_name|state_abbr| state_capital|state_id|
+--------------+----------+--------------+--------+
| Alabama| AL| Montgomery| 0|
| Alaska| AK| Juneau| 1|
| Arizona| AZ| Phoenix| 2|
| Arkansas| AR| Little Rock| 3|
| California| CA| Sacramento| 4|
| Colorado| CO| Denver| 5|
| Connecticut| CT| Hartford| 6|
| Delaware| DE| Dover| 7|
| Florida| FL| Tallahassee| 8|
| Georgia| GA| Atlanta| 9|
| Hawaii| HI| Honolulu| 10|
| Idaho| ID| Boise| 11|
| Illinois| IL| Springfield| 12|
| Indiana| IN| Indianapolis| 13|
| Iowa| IA| Des Moines| 14|
| Kansas| KS| Topeka| 15|
| Kentucky| KY| Frankfort| 16|
| Louisiana| LA| Baton Rouge| 17|
| Maine| ME| Augusta| 18|
| Maryland| MD| Annapolis| 19|
| Massachusetts| MA| Boston| 20|
| Michigan| MI| Lansing| 21|
| Minnesota| MN| St. Paul| 22|
| Mississippi| MS| Jackson| 23|
| Missouri| MO|Jefferson City| 24|
| Montana| MT| Helena| 25|
| Nebraska| NE| Lincoln| 26|
| Nevada| NV| Carson City| 27|
| New Hampshire| NH| Concord| 28|
| New Jersey| NJ| Trenton| 29|
| New Mexico| NM| Santa Fe| 30|
| New York| NY| Albany| 31|
|North Carolina| NC| Raleigh| 32|
| North Dakota| ND| Bismarck| 33|
| Ohio| OH| Columbus| 34|
| Oklahoma| OK| Oklahoma City| 35|
| Oregon| OR| Salem| 36|
| Pennsylvania| PA| Harrisburg| 37|
| Rhode Island| RI| Providence| 38|
|South Carolina| SC| Columbia| 39|
| South Dakota| SD| Pierre| 40|
| Tennessee| TN| Nashville| 41|
| Texas| TX| Austin| 42|
| Utah| UT|Salt Lake City| 43|
| Vermont| VT| Montpelier| 44|
| Virginia| VA| Richmond| 45|
| Washington| WA| Olympia| 46|
| West Virginia| WV| Charleston| 47|
| Wisconsin| WI| Madison| 48|
| Wyoming| WY| Cheyenne| 49|
+--------------+----------+--------------+--------+monotonically_increasing_id – two partition dataframe
scala> df_usa2.withColumn("state_id",monotonically_increasing_id).show(50)
+--------------+----------+--------------+----------+
| state_name|state_abbr| state_capital| state_id|
+--------------+----------+--------------+----------+
| Alabama| AL| Montgomery| 0|
| Arizona| AZ| Phoenix| 1|
| California| CA| Sacramento| 2|
| Connecticut| CT| Hartford| 3|
| Florida| FL| Tallahassee| 4|
| Hawaii| HI| Honolulu| 5|
| Illinois| IL| Springfield| 6|
| Iowa| IA| Des Moines| 7|
| Kentucky| KY| Frankfort| 8|
| Maine| ME| Augusta| 9|
| Massachusetts| MA| Boston| 10|
| Minnesota| MN| St. Paul| 11|
| Missouri| MO|Jefferson City| 12|
| Montana| MT| Helena| 13|
| Nevada| NV| Carson City| 14|
| New Jersey| NJ| Trenton| 15|
| New York| NY| Albany| 16|
| North Dakota| ND| Bismarck| 17|
| Oklahoma| OK| Oklahoma City| 18|
| Pennsylvania| PA| Harrisburg| 19|
|South Carolina| SC| Columbia| 20|
| Tennessee| TN| Nashville| 21|
| Utah| UT|Salt Lake City| 22|
| Virginia| VA| Richmond| 23|
| West Virginia| WV| Charleston| 24|
| Wyoming| WY| Cheyenne| 25|
| Alaska| AK| Juneau|8589934592|
| Arkansas| AR| Little Rock|8589934593|
| Colorado| CO| Denver|8589934594|
| Delaware| DE| Dover|8589934595|
| Georgia| GA| Atlanta|8589934596|
| Idaho| ID| Boise|8589934597|
| Indiana| IN| Indianapolis|8589934598|
| Kansas| KS| Topeka|8589934599|
| Louisiana| LA| Baton Rouge|8589934600|
| Maryland| MD| Annapolis|8589934601|
| Michigan| MI| Lansing|8589934602|
| Mississippi| MS| Jackson|8589934603|
| Nebraska| NE| Lincoln|8589934604|
| New Hampshire| NH| Concord|8589934605|
| New Mexico| NM| Santa Fe|8589934606|
|North Carolina| NC| Raleigh|8589934607|
| Ohio| OH| Columbus|8589934608|
| Oregon| OR| Salem|8589934609|
| Rhode Island| RI| Providence|8589934610|
| South Dakota| SD| Pierre|8589934611|
| Texas| TX| Austin|8589934612|
| Vermont| VT| Montpelier|8589934613|
| Washington| WA| Olympia|8589934614|
| Wisconsin| WI| Madison|8589934615|
+--------------+----------+--------------+----------+monotonically_increasing_id – three partition dataframe
scala> df_usa3.withColumn("state_id",monotonically_increasing_id).show(50)
+--------------+----------+--------------+-----------+
| state_name|state_abbr| state_capital| state_id|
+--------------+----------+--------------+-----------+
| Arizona| AZ| Phoenix| 0|
| Colorado| CO| Denver| 1|
| Florida| FL| Tallahassee| 2|
| Idaho| ID| Boise| 3|
| Iowa| IA| Des Moines| 4|
| Louisiana| LA| Baton Rouge| 5|
| Massachusetts| MA| Boston| 6|
| Mississippi| MS| Jackson| 7|
| Nevada| NV| Carson City| 8|
| New Mexico| NM| Santa Fe| 9|
| North Dakota| ND| Bismarck| 10|
| Oregon| OR| Salem| 11|
|South Carolina| SC| Columbia| 12|
| Texas| TX| Austin| 13|
| Virginia| VA| Richmond| 14|
| Wisconsin| WI| Madison| 15|
| Alabama| AL| Montgomery| 8589934592|
| Arkansas| AR| Little Rock| 8589934593|
| Connecticut| CT| Hartford| 8589934594|
| Georgia| GA| Atlanta| 8589934595|
| Illinois| IL| Springfield| 8589934596|
| Kansas| KS| Topeka| 8589934597|
| Maine| ME| Augusta| 8589934598|
| Michigan| MI| Lansing| 8589934599|
| Missouri| MO|Jefferson City| 8589934600|
| Montana| MT| Helena| 8589934601|
| New Hampshire| NH| Concord| 8589934602|
| New York| NY| Albany| 8589934603|
| Ohio| OH| Columbus| 8589934604|
| Pennsylvania| PA| Harrisburg| 8589934605|
| South Dakota| SD| Pierre| 8589934606|
| Utah| UT|Salt Lake City| 8589934607|
| Washington| WA| Olympia| 8589934608|
| Wyoming| WY| Cheyenne| 8589934609|
| Alaska| AK| Juneau|17179869184|
| California| CA| Sacramento|17179869185|
| Delaware| DE| Dover|17179869186|
| Hawaii| HI| Honolulu|17179869187|
| Indiana| IN| Indianapolis|17179869188|
| Kentucky| KY| Frankfort|17179869189|
| Maryland| MD| Annapolis|17179869190|
| Minnesota| MN| St. Paul|17179869191|
| Nebraska| NE| Lincoln|17179869192|
| New Jersey| NJ| Trenton|17179869193|
|North Carolina| NC| Raleigh|17179869194|
| Oklahoma| OK| Oklahoma City|17179869195|
| Rhode Island| RI| Providence|17179869196|
| Tennessee| TN| Nashville|17179869197|
| Vermont| VT| Montpelier|17179869198|
| West Virginia| WV| Charleston|17179869199|
+--------------+----------+--------------+-----------+Important points for monotonically_increasing_id
- For dataframe with 1 partition, all numbers are generated in sequence without any gaps
- For dataframe with 2 partition, all numbers are generated in sequence without any gaps for each partition with definite offset.
- For dataframe with 3 partition, all numbers are generated in sequence without any gaps for each partition with definite offset.
- There is significant difference between generated numbers when partition is changed i.e. offset is a big number.
- The 3 partitions are created with almost same number of rows in each partition. The exchange of data happens using RoundRobinPartitioning technique which takes care of even distribution of data across partitions.
- The offset is 2 power 33 i.e. 8589934592. So for every partition the offset will be big number * partition number. Like 1st partition starts from 85899345920=0 ; 2nd partition starts from: 85899345921 = 8589934592; 3rd partition starts from : 8589934592*2=17179869184 and so on.
- We can also conclude from above point that this method assumes that a partition will not have more than 8589934592 records i.e. ~8 Bn records in a single partition.
- Moving all data into a single partition can result in memory issues if volume is too high.
