Change pbix data source programmatically

Got it I think.

Here's Python code to unpack, change and repack the *.pbit programmatically (haven't tested on *.pbix):

----

import zipfile
import io

main_pbit = r"target_full_path_to_pbit/file.pbit"

revised_pbit =   r"target_full_path_to_save_pbit/file_new.pbit"


'''
Key Functions
'''    
def edit_Section1(old_text,new_text,datamashup_byte_stream):
    byte_dict = {}
    
    with datamashup_byte_stream as f:
        f.seek(0)
        #parsed bytes based on the microsoft MS-QDEFF documentation (thank you MS!)
        byte_dict['version'] = f.read(4)
        
        byte_dict['pkg_parts_len'] = f.read(4)
        byte_dict['pkg_parts'] = f.read(int.from_bytes(byte_dict['pkg_parts_len'],byteorder='little'))
        
        byte_dict['perm_len'] = f.read(4)
        byte_dict['perm_var'] = f.read(int.from_bytes(byte_dict['perm_len'],byteorder='little'))
        
        byte_dict['meta_len'] = f.read(4)
        byte_dict['meta_var'] = f.read(int.from_bytes(byte_dict['meta_len'],byteorder='little'))
        
        byte_dict['perm_bind_len'] = f.read(4)
        byte_dict['perm_bind_var'] = f.read(int.from_bytes(byte_dict['perm_bind_len'],byteorder='little'))
        
        mashup_f = io.BytesIO(byte_dict['pkg_parts'] )
        test_zip = zipfile.ZipFile(mashup_f,mode='r')
        archive_contents = test_zip.filelist
        #step into the package parts of the archive, which contains the query definitions
        with test_zip.open('Formulas/Section1.m','r') as section1_f:
            section1_text = section1_f.read()
            '''
            Do all your text swapping/manipulation here
            '''
            section1_revised = section1_text.replace(old_text,new_text)
        
        
        #create a new zip file in memory
        zip_buffer = io.BytesIO()
        zip_archive = zipfile.ZipFile(zip_buffer,mode='w',compression=zipfile.ZIP_DEFLATED)
        
        for ac in archive_contents:
            if ac.filename == 'Formulas/Section1.m':
                zip_archive.writestr('Formulas/Section1.m',data=section1_revised)
            else:
                with test_zip.open(ac,'r') as temp_f:
                    temp_bytes = temp_f.read()
                zip_archive.writestr(ac,data=temp_bytes)
        
        #close out streams   
        zip_archive.close()
        test_zip.close()
        
        #go back to start of bytestream and re-read it, finding new package parts and package parts length
        zip_buffer.seek(0)
        byte_dict['pkg_parts'] = zip_buffer.read()       
        byte_dict['pkg_parts_len'] = len(byte_dict['pkg_parts']).to_bytes(4,byteorder='little')
        
        f.close()
    
    new_mashup = io.BytesIO()
    
    #write out the revised DataMashup file (in correct order, from ordered list)
    for b in ['version','pkg_parts_len','pkg_parts','perm_len','perm_var','meta_len','meta_var','perm_bind_len','perm_bind_var']:
        new_mashup.write(byte_dict[b])
        
    return new_mashup
    

'''
Main Code
'''

main_zip = zipfile.ZipFile(main_pbit,mode='r')
main_archive_list = main_zip.filelist

new_zipfile = zipfile.ZipFile(revised_pbit,mode='w',compression=zipfile.ZIP_DEFLATED)

for al in main_archive_list:
    if al.filename == 'DataMashup':
        with main_zip.open(al.filename,'r') as datamashup_byte_stream:
            #Do any text parsing on the source definitions here
            revised_DataMashup = edit_Section1(b'CHANGETHISTOSOMETHINGELSE',b'Hello_world',datamashup_byte_stream)
            revised_DataMashup.seek(0)
            final_DataMashup = revised_DataMashup.read()
            new_zipfile.writestr('DataMashup',data=final_DataMashup)
    else:
        with main_zip.open(al,'r') as temp_f:
            temp_bytes = temp_f.read()
        new_zipfile.writestr(al,data=temp_bytes)
        
#close out streams   
main_zip.close()
new_zipfile.close()

----

For this demo I just made a very basic pbit file with a data source and a custom column:

let
     Source = ...

     ...
     #"Added Custom" = Table.AddColumn(#"Removed Other Columns", "Custom_Col", each "CHANGETHISTOSOMETHINGELSE")
in
     #"Added Custom"

... and it looks like we can swap out the column text "CHANGETHISTOSOMETHINGELSE"-->"Hello_World" fine.  To swap out other parts of the M query source, I think it's just a matter of the specific text parsing for section1.m 

Next step will be to figure out how to pull/push these files to/from webservice, but hopefully this is a useful piece for others.