Python 2.7 cmd自动完成读取行缓冲区似乎过时

Question

I'm using readline based cmd module on Mac OS (and Red Hat Linux) with tab completion enabled. It seems to be working great except for when I introduced my own logic to handle "sub command" completions based on regex matching. here is a simple address book example where I have a tree based structure:

Code:

import cmd
import readline
import re

readline.parse_and_bind("bind ^I rl_complete")

class Test(cmd.Cmd):
    def do_assign(self, line):
        print line
    def complete_assign(self, text, line, begidx, endidx):
        if re.match('assign\s+employee\s+name', line):
            return [i for i in ['first', 'middle', 'last']
                if i.startswith(text)]
        elif re.match('assign\s+employee', line):
            return [i for i in ['name', 'address', 'phone']
                if i.startswith(text)]
        elif line.startswith("assign"):
            return [i for i in ['employee', 'manager']
                if i.startswith(text)]

Test().cmdloop()

Output:

(Cmd) assign <TAB>
employee  manager   

(Cmd) assign employee <TAB>
address  phone    name 

(Cmd) assign employee name <TAB>
last    middle  first   

(Cmd) assign employee name first <RETURN>
employee name first

(Cmd) assign <TAB> <-- This is incorrect. Expecting ['employee', 'manager'] instead
last    middle  first  

(Cmd) assign

As you see the difference between first and last "assign " operations, I'm getting different results because the line variable in function complete_assign seems to get assigned assign employee name first instead of just assign because of the previous RETURN operation, due to which the completor logic is now matching the wrong regex code block and providing the incorrect completions. Is there a way to "flush" the readline stdin buffer after every operation so that it doesn't remember the buffer from previous complete_assign or do_assign trials?

Thanks!

Answer 1

I think I solved it, thanks to this article: http://pymotw.com/2/readline/#accessing-the-completion-buffer

While going through the BufferAwareCompleter code, I stumbled across this line:

being_completed = origline[begin:end]

Which made me think that I should be able to use the line indices provided in the completer API to my advantage, and this is exactly what helped.

Since I did not care about the 'stale' part of the buffer, I just used end index to demarcate my line boundary. This did the trick:

line = line[:endidx]

Updated working code:

import cmd
import readline
import re

readline.parse_and_bind("bind ^I rl_complete")

class Test(cmd.Cmd):
    def do_assign(self, line):
        print line
    def complete_assign(self, text, line, begidx, endidx):
        line = line[:endidx] # <-- ADDED THIS
        if re.match('assign\s+employee\s+name', line):
            return [i for i in ['first', 'middle', 'last']
                if i.startswith(text)]
        elif re.match('assign\s+employee', line):
            return [i for i in ['name', 'address', 'phone']
                if i.startswith(text)]
        elif line.startswith("assign"):
            return [i for i in ['employee', 'manager']
                if i.startswith(text)]

Test().cmdloop()

Update working output:

(Cmd) assign 
employee  manager   
(Cmd) assign employee 
name     phone    address  
(Cmd) assign employee name 
last    middle  first   
(Cmd) assign employee name first
employee name first
(Cmd) assign 
manager   employee  
(Cmd) assign