I am trying to join multiple dataframes in PySpark by one chained operation. The join key column name is the same in all of them. The code snippet:
columns_summed = [i for i in df_summed.columns if i != "buildingBlock_id"]
columns_concat = [i for i in df_concat.columns if i != "buildingBlock_id"]
columns_indicator = [i for i in df_indicator_fields.columns if i != "buildingBlock_id"]
columns_takeone = [i for i in df_takeone.columns if i != "buildingBlock_id"]
columns_minmax = [i for i in df_minmax.columns if i != "buildingBlock_id"]
df_all_joined = (df_summed.alias("df1").join(df_concat,df_summed.buildingBlock_id == df_concat.buildingBlock_id, "left")
.join(df_indicator_fields,df_summed.buildingBlock_id == df_indicator_fields.buildingBlock_id, "left")
.join(df_takeone,df_summed.buildingBlock_id == df_takeone.buildingBlock_id, "left")
.join(df_minmax,df_summed.buildingBlock_id == df_minmax.buildingBlock_id, "left")
.select("df1.buildingBlock_id", *columns_summed
, *columns_concat
, *columns_indicator
, *columns_takeone
, *columns_minmax
)
)
Now, when I am trying to display the joined dataframe using:
display(df_all_joined)
I'm getting the following error:
AnalysisException: Reference 'df1.buildingBlock_id' is ambiguous, could be: df1.buildingBlock_id, df1.buildingBlock_id.
Why am I getting this error even though I specified where the key column should come from?
You should specify the join columns as an array of strings:
.join(df_concat,['buildingBlock_id'], "left")
If the columns that you are joining on have the same name, this makes sure to drop one of them. In the case of the left join, it drops the column from df_concat.
If you don't do that, you end up with both columns in the joined data frame, thus creating this "Ambiguous" excception.
Related
Python doesn't like the ampersand below.
I get the error: & is not a supported operation for types str and str. Please review your code.
Any idea how to get this right? I've never tried to join more than 1 column for aliased tables. Thx!!
df_initial_sample = df_crm.alias('crm').join(df_cngpt.alias('cng'), on= (("crm.id=cng.id") & ("crm.cpid = cng.cpid")), how = "inner")
Try using as below -
df_initial_sample = df_crm.alias('crm').join(df_cngpt.alias('cng'), on= (["id"] and ["cpid"]), how = "inner")
Your join condition is overcomplicated. It can be as simple as this
df_initial_sample = df_crm.join(df_cngpt, on=['id', 'cpid'], how = 'inner')
My probleme is i have a code that gives filter column and values in a list as parameters
val vars = "age IN ('0')"
val ListPar = "entered_user,2014-05-05,2016-10-10;"
//val ListPar2 = "entered_user,2014-05-05,2016-10-10;revenue,0,5;"
val ListParser : List[String] = ListPar.split(";").map(_.trim).toList
val myInnerList : List[String] = ListParser(0).split(",").map(_.trim).toList
if (myInnerList(0) == "entered_user" || myInnerList(0) == "date" || myInnerList(0) == "dt_action"){
responses.filter(vars +" AND " + responses(myInnerList(0)).between(myInnerList(1), myInnerList(2)))
}else{
responses.filter(vars +" AND " + responses(myInnerList(0)).between(myInnerList(1).toInt, myInnerList(2).toInt))
}
well for all the fields except the one that contains date the functions works flawless but for fields that have date it throws an error
Note : I'm working with parquet files
here is the error
when i try to write it manually i get the same
here is how the query it sent to the sparkSQL
the first one where there is revenue it works but the second one doesn't work
and when i try to just filter with dates without the value of "vars" which contains other columns, it works
Well my issue is that i was mixing between sql and spark and when i tried to concatenate sql query which is my variable "vars" whith df.filter() and especially when i used between operator it was giving an output format unrocognised by sparksql which is
age IN ('0') AND ((entered_user >= 2015-01-01) AND (entered_user <= 2015-05-01))
it might seems correct but after looking in sql documentation it was missing parenthesese(in vars) it needed to be
(age IN ('0')) AND ((entered_user >= 2015-01-01) AND (entered_user <= 2015-05-01))
well the solution is i needed to concatenate those correctly so to do that i must to add " expr " to the variable vars which will result the desire syntaxe
responses.filter(expr(vars) && responses(myInnerList(0)).between(myInnerList(1), myInnerList(2)))
I have a DF A_DF which has among others two columns say COND_B and COND_C. Then I have 2 different df's B_DF with COND_B column and C_DF with COND_C column.
Now I would like to filter A_DF where the value match in one OR the other. Something like:
df = A_DF.filter((A_DF.COND_B == B_DF.COND_B) | (A_DF.COND_C == C_DF.COND_C))
But I found out it is not possible like this.
EDIT
error: Attribute CON_B#264,COND_C#6 is missing from the schema: [... COND_B#532, COND_C#541 ]. Attribute(s) with the same name appear in the operation: COND_B,COND_C. Please check if the right attribute(s) are used.; looks like I can filter only on same DF because of the #number added on the fly..
So I first tried to do a list from B_DF and C_DF and use filter based on that but it was too expensive to use collect() on 100m of records.
So I tried:
AB_DF = A_DF.join(B_DF, 'COND_B', 'left_semi')
AC_DF = A_DF.join(C_DF, 'COND_C', 'left_semi')
df = AB_DF.unionAll(AC_DF).dropDuplicates()
dropDuplicates() I used to removed duplicate records where both conditions where true. But even with that I got some unexpected results.
Is there some other - smoother solution to do it simply? Something like an EXISTS statement in SQL?
EDIT2
I tried SQL based on #mck response:
e.createOrReplaceTempView('E')
b.createOrReplaceTempView('B')
p.createOrReplaceTempView('P')
df = spark.sql("""select * from E where exists (select 1 from B where E.BUSIPKEY = B.BUSIPKEY) or exists (select 1 from P where E.PCKEY = P.PCKEY)""")
my_output.write_dataframe(df)
with error:
Traceback (most recent call last):
File "/myproject/abc.py", line 45, in my_compute_function
df = spark.sql("""select * from E where exists (select 1 from B where E.BUSIPKEY = B.BUSIPKEY) or exists (select 1 from P where E.PCKEY = P.PCKEY)""")
TypeError: sql() missing 1 required positional argument: 'sqlQuery'
Thanks a lot!
Your idea of using exists should work. You can do:
A_DF.createOrReplaceTempView('A')
B_DF.createOrReplaceTempView('B')
C_DF.createOrReplaceTempView('C')
df = spark.sql("""
select * from A
where exists (select 1 from B where A.COND_B = B.COND_B)
or exists (select 1 from C where A.COND_C = C.COND_C)
""")
I'm not sure why this is happening. In PySpark, I read in two dataframes and print out their column names and they are as expected, but then when do a SQL join I get an error that cannot resolve column name given the inputs. I have simplified the merge just to get it to work, but I will need to add in more join conditions which is why I'm using SQL (will be adding in: "and b.mnvr_bgn < a.idx_trip_id and b.mnvr_end > a.idx_trip_data"). It appears that the column 'device_id' is being renamed to '_col7' in the df mnvr_temp_idx_prev_temp
mnvr_temp_idx_prev = mnvr_3.select('device_id', 'mnvr_bgn', 'mnvr_end')
print mnvr_temp_idx_prev.columns
['device_id', 'mnvr_bgn', 'mnvr_end']
raw_data_filtered = raw_data.select('device_id', 'trip_id', 'idx').groupby('device_id', 'trip_id').agg(F.max('idx').alias('idx_trip_end'))
print raw_data_filtered.columns
['device_id', 'trip_id', 'idx_trip_end']
raw_data_filtered.registerTempTable('raw_data_filtered_temp')
mnvr_temp_idx_prev.registerTempTable('mnvr_temp_idx_prev_temp')
test = sqlContext.sql('SELECT a.device_id, a.idx_trip_end, b.mnvr_bgn, b.mnvr_end \
FROM raw_data_filtered_temp as a \
INNER JOIN mnvr_temp_idx_prev_temp as b \
ON a.device_id = b.device_id')
Traceback (most recent call last): AnalysisException: u"cannot resolve 'b.device_id' given input columns: [_col7, trip_id, device_id, mnvr_end, mnvr_bgn, idx_trip_end]; line 1 pos 237"
Any help is appreciated!
I would recommend renaming the name of the field 'device_id' in at least one of the data frame. I modified your query just a bit and tested it(in scala). Below query works
test = sqlContext.sql("select * FROM raw_data_filtered_temp a INNER JOIN mnvr_temp_idx_prev_temp b ON a.device_id = b.device_id")
[device_id: string, mnvr_bgn: string, mnvr_end: string, device_id: string, trip_id: string, idx_trip_end: string]
Now if you are doing a 'select * ' in above statement, it will work. But if you try to select 'device_id', you will get an error "Reference 'device_id' is ambiguous" . As you can see in the above 'test' data frame definition, it has two fields with the same name(device_id). So to avoid this, I recommend changing field name in one of the dataframes.
mnvr_temp_idx_prev = mnvr_3.select('device_id', 'mnvr_bgn', 'mnvr_end')
.withColumnRenamned("device_id","device")
raw_data_filtered = raw_data.select('device_id', 'trip_id', 'idx').groupby('device_id', 'trip_id').agg(F.max('idx').alias('idx_trip_end'))
Now use dataframes or sqlContext
//using dataframes with multiple conditions
val test = mnvr_temp_idx_prev.join(raw_data_filtered,$"device" === $"device_id"
&& $"mnvr_bgn" < $"idx_trip_id","inner")
//in SQL Context
test = sqlContext.sql("select * FROM raw_data_filtered_temp a INNER JOIN mnvr_temp_idx_prev_temp b ON a.device_id = b.device and a. idx_trip_id < b.mnvr_bgn")
Above queries will work for your problem. And if your data set is too large, I would recommend to not use '>' or '<' operators in Join condition as it causes cross join which is a costly operation if data set is large. Instead use them in WHERE condition.
Can someone please tell me how do we compare a table column against NULL using slick's lifted embedding.
What i want to achieve in mysql:
select * from Users where email = '<some_email_addr>' and ( removed = NULL OR removed <= 1 )
It gives me an errors at x.removed === null when i try:
val q = for {
x <- Users
if x.email === email && ( x.removed === null || x.removed <= 1 )
} yield x.removed
Thanks
Try:
x.removed.isNull
I think that is what you are looking for
As daaatz said and per http://slick.typesafe.com/doc/2.1.0/upgrade.html#isnull-and-isnotnull you should use isEmpty, but it only works on Option columns. A workaround is x.removed.?.isEmpty.
Try
x.removed.isEmpty()
There are no nulls in scala ;-)