Conditionally merge columns using Pandas

Question

[frcamacho]

I start out with this pandas dataframe:

sampleID    scaffoldID    Type   Program    Breadth  \
3   G38791    scaffold_7     4  A  73.558964  
0   G38791  scaffold_388     3      B   0.000000  
1   G38791  scaffold_777     2      B   0.000000  
2   G38791  scaffold_787     0      B   0.000000  
3   G38791    scaffold_7     4      B  73.558964

How can I conditionally merge columns? So if df['Type' ==4], I want to change Type value for that row to "Partial" then merge column value at Program and Breadth value to give a new value for the column, Type to partial_A_73.558964?

New dataframe should be:

 sampleID    scaffoldID    Type   Program    Breadth  \
3   G38791    scaffold_7     partial_A_73.558964  A  73.558964   
0   G38791  scaffold_388     3      B   0.000000   
1   G38791  scaffold_777     2      B   0.000000   
2   G38791  scaffold_787     0      B   0.000000   
3   G38791    scaffold_7     partial_B_73.558964 B  73.558964

Answer 1

[frcamacho]

This code works:

df['Type'] = np.where(df['Type'] == 4, "partial"+df['Program']+"_"+df['Breadth'],df['Type'])

Other solutions are welcome!

[Ram]

Quick question - I think mixing data types in a column is a bad idea. Is pursuing an alternative that maintains data integrity a viable option?

[frcamacho]

I will be dropping both Program and Breadth column which is why I needed to concat the rows. I agree with you that in general you should not mix data types but, because of the analysis I will be conducting the Type column tell us a lot about the row, especially those that are partials.

[Ram]

Hmmm. It's just a matter of personal choice, I guess. Personally, in a storage vs maintainability contest, I'd pick maintainability (especially when atomicity is at stake), but that's me being excessively obsessed with data integrity.