Convert a dataframe to a list of tuples

Question

I have a table pandas DF which looks like

	Slave	start_addr0	end_addr0	start_addr1	end_addr1	start_addr2	end_addr2
0	0	10000000	1FFFFFFF	NaN	NaN	NaN	NaN
1	1	20000000	2007FFFF	40000000	40005FFF	NaN	NaN
2	1	20000000	2007FFFF	20100000	201FFFFF	NaN	NaN
3	2	20200000	202FFFFF	20080000	20085FFF	40006000	400FFFFF
4	3	0	0FFFFFFF	NaN	NaN	NaN	NaN
5	4	20300000	203FFFFF	NaN	NaN	NaN	NaN
6	5	20400000	204FFFFF	NaN	NaN	NaN	NaN

For each slave number I need to convert it to a list of ranges (tuples). For example,

Slave1_list = ( (20000000, 2007FFFF), (40000000, 40005FFF), (20100000, 201FFFFF))

The number of slaves (rows) and address-pairs (columns) can vary.

Thanks

EDIT:

Run the following code to load sample data into dataframe:

import pandas as pd
import io

f = io.StringIO('''Slave|start_addr0|end_addr0|start_addr1|end_addr1|start_addr2|end_addr2
0|10000000|1FFFFFFF|NaN|NaN|NaN|NaN
1|20000000|2007FFFF|40000000|40005FFF|NaN|NaN
1|20000000|2007FFFF|20100000|201FFFFF|NaN|NaN
2|20200000|202FFFFF|20080000|20085FFF|40006000|400FFFFF
3|0|0FFFFFFF|NaN|NaN|NaN|NaN
4|20300000|203FFFFF|NaN|NaN|NaN|NaN
5|20400000|204FFFFF|NaN|NaN|NaN|NaN
''')
df = pd.read_csv(f, sep='|', engine='python', index_col=None)

Answer 1

You can try:

One option via wide_to_long:

df = df.reset_index()
result = pd.wide_to_long(df, stubnames=['start_addr', 'end_addr'], i=['index', 'Slave'], j='add_num', sep='').dropna(
).reset_index([0, -1], drop=True).apply(tuple, 1).groupby(level=0).agg(list)

An option via groupby:

k = df.set_index('Slave').stack().reset_index()
result = k.groupby(k.index//2).agg({'Slave': 'first', 0 : tuple}).groupby('Slave').agg({0 : set})

Explanation:

df.set_index('Slave').stack().reset_index() will remove the NaN values and stack the dataframe.

k.groupby(k.index//2) will group alternate rows and perform the required aggregations(tuples are formed in this step)

.groupby('Slave').agg({0 : set}) -> Last groupby is to capture the unique tuples for each slave.

OUTPUT:

                                                                            0
Slave                                                                        
0                                                      {(10000000, 1FFFFFFF)}
1      {(40000000.0, 40005FFF), (20100000.0, 201FFFFF), (20000000, 2007FFFF)}
2      {(20080000.0, 20085FFF), (40006000.0, 400FFFFF), (20200000, 202FFFFF)}
3                                                             {(0, 0FFFFFFF)}
4                                                      {(20300000, 203FFFFF)}
5                                                      {(20400000, 204FFFFF)}

NOTE: I'm assuming for every start_addr there exists an end_addr.