I've been learning pandas for a couple of days. I am migrating a SQL DB to PYTHON and have encountered the sql statement (example):
select * from
table_A a
left join table_B b
on a.ide = b.ide
and a.credit_type = case when b.type > 0 then b.credit_type else a.credit_type end
I've only been able to migrate to the first condition. My difficulty is in the last line and I don't know how to migrate it. Tables are actually sql queries that I've stored in dataframes.
merge = pd.merge(df_query_a, df_query_b),on='ide', how='left')
any suggestions please.
The Case condition is like an if-then-else statement, which you can implement in Pandas using np.where() like below:
Based on left join resulting dataframe merge:
import numpy as np
merge['credit_type_x'] = np.where(merge['type_y'] > 0, merge['credit_type_y'], merge['credit_type_x'])
Here the column names credit_type_x credit_type_y should have been created on the Pandas merge function after renaming conflicting (same) column names on the 2 sources dataframes. In case dataframe merge doesn't have the column type_y because column type appears only on Table_B but not on Table_A, you can use column name type here instead.
Alternatively, as you just need to modify the value of credit_type_x only when type_y > 0 and retain the value of credit_type_x without modification if NOT type_y > 0, we can also do it simply by:
merge.loc[merge['type_y'] > 0, 'credit_type_x'] = merge['credit_type_y']
Below two options to face your problem
You can add a column in df_query_a based on the condition that you need considering two dataframes, and after that, make the merge.
You can try with some library as pandasql3.
Related
I have a delta table in Databricks that I am loading into a dataframe to perform a specific numpy operation:
import pyspark.pandas as ps
df = spark.sql("SELECT Id,Field_to_Transform FROM bronze.Table_A where Field_to_Transform is not null").toPandas()
The operation I perform is to remove special characters:
df['Transformed_Field'] = [''.join(e for e in x if e.isalnum()) for x in df['Field_to_Transform ']]
df.drop(['Field_to_Transform '], axis=1)
So this leaves me with the dataframe "df" which has just the id and the Transformed_Field in it:
Id
Transformed_Field
00A1
12345
00A2
123456
00A3
1234
00A4
1234568
Now I want to left join the df back to bronze.Table_A in databricks by simply joining back on the id field.
What is the most effecient way to join df back to bronze.Table_A?
Things I have tried so far:
Saved the entire bronze.Table_A in a new dataframe df2, used df.merge to put them together, and then created a brand new table. This worked but it was way too excessive and I do not need a new table, just the transformed column joined back.
Tried to use spark.sql to perform the left join in a SQL query but it seems I cannot access a df inside a spark.sql query.
Any help is much appreciated, thank you.
Option 1 - DataFrame API way
The first option is a modification of your first bullet point:
Saved the entire bronze.Table_A in a new dataframe df2, used df.merge to put them together, and then created a brand new table. This worked but it was way too excessive and I do not need a new table, just the transformed column joined back.
The merge operation is a Pandas method and not PySpark, hence the performance might not be optimal - docs here. Loading the whole table into DataFrame is the correct approach, that just needs the built-in join method to the required left-join, like this:
table_df = spark.read.table("bronze.Table_A")
# Join
merged_df = table_df.join(df, on="Id", how="left")
Option 2 - SQL way
The second option builds on your second bullet point:
Tried to use spark.sql to perform the left join in a SQL query but it seems I cannot access a df inside a spark.sql query.
You can temporarily register DataFrame as a view, and then query it using plain SQL, whether in spark.sql method or any other way. Try doing this:
df.createOrReplaceTempView("transformed_df")
# Then join like this
spark.sql("SELECT * FROM bronze.Table_A ta LEFT JOIN transformed_df tdf ON ta.Id = tdf.Id")
I'm working on how to cluster a patstat (reference database) database.
With my own agorithm I came up with a dataframe which shows me the author, beginpage, endpage, volume and publication_year of a reference.
running:
dfhead = df.head(10)
shows me
Now I want the following:
Show inner join with the SAME dataframe such that for example author, beginpage and endpage are the same. (at least 3 similarities between the rows)
I tried:
c = ['author', 'beginpage','endpage', 'volume','publication year']
df_merge = dfhead.merge(dfhead, how = 'inner',on = [c[0],c[1],c[2]])
where
The answer will then be given such that there only exists an inner join with exactly the same row, but I don't want those to include.
In the example above the df_merge should not take any values since there are no 3 similar columns.
What if there would be some how to same row, I will show an example:
x = pd.Dataframe({'author':['lee','lee'], 'beginpage':[455,456],'endpage':[477,477],'volume':[300,300]})
Note that the two rows have (at least) 3 similar columns and therefore the merge/join should be visible.
BUT note that in don't want to include to join of exactly the same row!!!
You could do an inner join and apply filtering to exclude the same row, but maybe it would be more straightforward to use groupby instead?
df.groupby(by=['author', 'beginpage','endpage'])
https://pandas.pydata.org/pandas-docs/stable/reference/api/pandas.DataFrame.groupby.html
Applying aggregations/calculations/ect to groups:
https://pandas.pydata.org/pandas-docs/stable/reference/groupby.html
So I have 2 tables, Table 1 and Table 2, Table 2 is sorted with the dates- recent dates to old dates. So in excel when I do a lookup in Table 1 and the lookup is done from Table 2, It only picks the first value from table 2 and does not move on to search for the same value after the first.
So I tried replicating it in python with the merge function, but found out it gets to repeat the value the number of times it appears in the second table.
pd.merge(Table1, Table2, left_on='Country', right_on='Country', how='left', indicator='indicator_column')
TABLE1
TABLE2
Merger result
Expected Result(Excel vlookup)
Is there any way this could be achieved with the merge function or any other python function?
Typing this in the blind as you are including your data as images, not text.
# The index is a very important element in a DataFrame
# We will see that in a bit
result = table1.set_index('Country')
# For each country, only keep the first row
tmp = table2.drop_duplicates(subset='Country').set_index('Country')
# When you assign one or more columns of a DataFrame to one or more columns of
# another DataFrame, the assignment is aligned based on the index of the two
# frames. This is the equivalence of VLOOKUP
result.loc[:, ['Age', 'Date']] = tmp[['Age', 'Date']]
result.reset_index(inplace=True)
Edit: Since you want a straight up Vlookup, just use join. It appears to find the very first one.
table1.join(table2, rsuffix='r', lsuffix='l')
The docs seem to indicate it performs similarly to a vlookup: https://pandas.pydata.org/pandas-docs/stable/reference/api/pandas.DataFrame.join.html
I'd recommend approaching this more like a SQL join than a Vlookup. Vlookup finds the first matching row, from top to bottom, which could be completely arbitrary depending on how you sort your table/array in excel. "True" database systems and their related functions are more detailed than this, for good reason.
In order to join only one row from the right table onto one row of the left table, you'll need some kind of aggregation or selection - So in your case, that'd be either MAX or MIN.
The question is, which column is more important? The date or age?
import pandas as pd
df1 = pd.DataFrame({
'Country':['GERM','LIB','ARG','BNG','LITH','GHAN'],
'Name':['Dave','Mike','Pete','Shirval','Kwasi','Delali']
})
df2 = pd.DataFrame({
'Country':['GERM','LIB','ARG','BNG','LITH','GHAN','LIB','ARG','BNG'],
'Age':[35,40,27,87,90,30,61,18,45],
'Date':['7/10/2020','7/9/2020','7/8/2020','7/7/2020','7/6/2020','7/5/2020','7/4/2020','7/3/2020','7/2/2020']
})
df1.set_index('Country')\
.join(
df2.groupby('Country')\
.agg({'Age':'max','Date':'max'}), how='left', lsuffix='l', rsuffix='r')
The drop duplicates methods of Spark DataFrames is not working and I think it is because the index column which was part of my dataset is being treated as a column of data. There definitely are duplicates in there, I checked it by comparing COUNT() and COUNT(DISTINCT()) on all the columns except the index. I'm new to Spark DataFrames but if I was using Pandas, at this point I would do pandas.DataFrame.set_index on that column.
Does anyone know how to handle this situation?
Secondly, there appears to be 2 methods on a Spark DataFrame, drop_duplicates and dropDuplicates. Are they the same?
If you don't want the index column to be considered while checking for the distinct records, you can drop the column using below command or select only the columns required.
df = df.drop('p_index') // Pass column name to be dropped
df = df.select('name', 'age') // Pass the required columns
drop_duplicates() is an alias for dropDuplicates().
https://spark.apache.org/docs/latest/api/python/pyspark.sql.html#pyspark.sql.DataFrame.dropDuplicates
What is the Pandas equivalent of this SQL code?
Select id, fname, lname from table where id = 123
I know that this is the equivalent of an SQL 'where' clause in Pandas:
df[df['id']==123]
And this selects specific columns:
df[['id','fname','lname']]
But I can't figure out how to combine them. All examples I've seen online select all columns with conditions. I want to select a limited number of columns with one or more conditions.
Use SQL-like .query() method:
df.query("id == 123")[['id','fname','lname']]
or
df[['id','fname','lname']].query("id == 123")
or more "Pandaic":
df.loc[df['id'] == 123, ['id','fname','lname']]
Extending on #MaxU's answer, suppose you needed multiple column values, taking 'fname'
df[['id','fname','lname']].query("fname == ('simon', 'michael')")
Without using query method of #MaxU, for simplicity included all columns:
df[df.fname.isin(['simon', 'michael'])]
Cascading the above with [['id','fname','lname']] will give the needed answer.