有没有更快的方法将大文件从hexa转换为二进制，二进制转换为int？

Question

I have a big DataFrame (1999048 rows and 1col), with hexadecimal datas. I want to put each line in binary, cut it into pieces and traduce each piece in decimal format.

I tried this:

for i in range (len(df.index)):
    hexa_line=hex2bin(str(f1.iloc[i]))[::-1] 
    channel = int(hexa_line[0:3][::-1], 2)     
    edge = int(hexa_line[3][::-1], 2)      
    time = int(hexa_line[4:32][::-1], 2)   
    sweep = int(hexa_line[32:48][::-1], 2)  
    tag = int(hexa_line[48:63][::-1], 2)   
    datalost = int(hexa_line[63][::-1], 2)   
    line=np.array([[channel, edge, time, sweep, tag, datalost]])
    tab=np.concatenate((tab, line), axis=0)

But it is really really long.... Is there a faster way to do that ?

Answer 1

only thing I can imagine helping a lot would be changing these lines:

line=np.array([[channel, edge, time, sweep, tag, datalost]])
tab=np.concatenate((tab, line), axis=0)

certainly in pandas, and I think also in numpy concatting is an expensive thing to do, and depends on the size of the total size of both arrays (rather than, say list.append)

I think what this does is re-writes the entire array tab each time you call it. Perhaps you could try appending each line to a list then concatting the whole list together.

eg something more like this:

tab = []
for i in range (len(df.index)):
    hexa_line=hex2bin(str(f1.iloc[i]))[::-1] 
    channel = int(hexa_line[0:3][::-1], 2)     
    edge = int(hexa_line[3][::-1], 2)      
    time = int(hexa_line[4:32][::-1], 2)   
    sweep = int(hexa_line[32:48][::-1], 2)  
    tag = int(hexa_line[48:63][::-1], 2)   
    datalost = int(hexa_line[63][::-1], 2)   
    line=np.array([[channel, edge, time, sweep, tag, datalost]])
    tab.append(line)

final_tab = np.concatenate(tab, axis=0)
# or whatever the syntax is :p