Merge lp://staging/~frankban/python-shelltoolbox/helpers into lp://staging/python-shelltoolbox

Thank you! Small comments and suggestions follow.

I asked about reordering the functions for alphabetization. Not having run at the top is somewhat surprising to me, but I am ok with it.

I was struck that file_append really expects the line to be added to end with a \n. I suppose we could enforce that one way or another...or not. I'm OK with leaving as is, or you could add a \n if the line does not already end with one (and do this before you check if the line is in the content). On a somewhat related note, it might be better to check if the line to be added is in the output of readlines(), not read(); that would verify that the line, not a line that ends with the line, is in the file. For example, right now a file of "abc\ndef\n" would incorrectly show that the "ef\n" line was in the file.

In file_prepend, I am tempted to suggest that this snippet should be changed:
            if line in lines:
                lines.remove(line)
This iterates through the lines twice, which is mildly annoying to me but almost certainly not a practical concern for us with these use cases. That said, I'd be tempted to change that to the following:
            try:
                lines.remove(line)
            except ValueError:
                pass
You could also verify that the line to be added in this function ended with a \n if you wanted to.

I don't really understand this part of the docstring for get_su_command:
    This can be used together with `run` when the `su` context manager is not
    enough (e.g. an external program uses uid rather than euid).
Oh! You mean that you would use run(*get_su_command(...)). Maybe an example would help.

I'm a bit concerned about what you are doing here in get_value_from_line:
    return line.split('=')[1].strip('"\' ')
would maybe json be a safer approach? No, I tried it, and json.loads('"hi"') is fine but json.loads("'hi'") is not. :-/ I dunno. When do we use this? The logic seems a bit idiosyncratic.

The grep command also seems a bit idiosyncratic. Why is it reasonable to always strip the result? I also would be tempted to rename the funtion: grep does a lot more than that. simply search_file? Hm, I see that you just reordered this file, you didn't actually add grep.... Never mind then, I guess. I should have commented on this sooner, but you shouldn't have to worry about it. If you *do* want to change it, please leave the "grep" alias around for now so that our charms do not suddenly break.

I don't understand why you removed the Serializer. Was it defined twice?

I thought the choices you made for doc examples and for unit tests were good. Thank you.

Great job!

Gary

> I was struck that file_append really expects the line to be added to end with
> a \n. I suppose we could enforce that one way or another...or not. I'm OK
> with leaving as is, or you could add a \n if the line does not already end
> with one (and do this before you check if the line is in the content). On a
> somewhat related note, it might be better to check if the line to be added is
> in the output of readlines(), not read(); that would verify that the line, not
> a line that ends with the line, is in the file. For example, right now a file
> of "abc\ndef\n" would incorrectly show that the "ef\n" line was in the file.

I agree with your suggestions. I've changed the function and removed those
ambiguities. I've added tests for line not ending with '\n' and for line
fragments.

> In file_prepend, I am tempted to suggest that this snippet should be changed:
> if line in lines:
> lines.remove(line)
> This iterates through the lines twice, which is mildly annoying to me but
> almost certainly not a practical concern for us with these use cases. That
> said, I'd be tempted to change that to the following:
> try:
> lines.remove(line)
> except ValueError:
> pass
> You could also verify that the line to be added in this function ended with a
> \n if you wanted to.

Done.

> I'm a bit concerned about what you are doing here in get_value_from_line:
> return line.split('=')[1].strip('"\' ')
> would maybe json be a safer approach? No, I tried it, and json.loads('"hi"')
> is fine but json.loads("'hi'") is not. :-/ I dunno. When do we use this?
> The logic seems a bit idiosyncratic.

I don't remember the origin of that function. Since it is not used in the
charms or in setuplxc/lpsetup, I am going to remove it.

> The grep command also seems a bit idiosyncratic. Why is it reasonable to
> always strip the result? I also would be tempted to rename the funtion: grep
> does a lot more than that. simply search_file? Hm, I see that you just
> reordered this file, you didn't actually add grep.... Never mind then, I
> guess. I should have commented on this sooner, but you shouldn't have to
> worry about it. If you *do* want to change it, please leave the "grep" alias
> around for now so that our charms do not suddenly break.

Actually that helper is not used, and I remember it was before we changed the
way the buildslave is reconfigured. I think it's safe to rename it and change
the returned value.

> I don't understand why you removed the Serializer. Was it defined twice?

Yes it was.

Thanks Gary.