Reinstate ordering of arguments in bin/storm.py

[sercuzz8]

Summary

argparse's parse_known_args() splits main_args (nargs='*') from unrecognized flags
(unknown_args), losing their relative order when they are interleaved on the command line.
This breaks down the flow of mvn clean install, thus the order must be reinstated.

Following this discussion

[GGraziadei]

Hi @sercuzz8,

Thanks for tackling this, the diagnosis is right, and reconstructing the order from sys.argv is the correct shape for the fix. A few things I'd like addressed before this goes in.

• The assignment replaces main_args instead of extending it. The previous line was main_args += unknown_args; the new one overwrites main_args with only the tokens found in argv. Anything in main_args that didn't come verbatim from the command line, an argparse default=[...] on a subparser positional, or values injected before main() reassembles, is now silently dropped. Could you confirm no subcommand relies on that?

• Tokens can be lost with no diagnostic.There's no else branch and no check that len(merged) == len(known_args) + len(unknown_args). If the invariant ever breaks, storm jar launches a topology with arguments missing and no error, it just looks like it worked. Either raise:

if len(merged) != len(known_args) + len(unknown_args):
    raise ValueError("could not reconstruct argument order from sys.argv")

or fall back to the old concatenation, if you'd rather not add a new failure path to a released CLI.

• Consider dropping the positional pointers. The two indices tie the function to both lists being ordered subsequences of argv. A count-based filter is shorter, order-agnostic, and gives you the leftover check for free:

from collections import Counter

remaining = Counter(known_args) + Counter(unknown_args)
merged = []
for token in sys_args:
    if remaining[token]:
        remaining[token] -= 1
        merged.append(token)

Nits: the #Recombine ... line should be a docstring with """

[sercuzz8]

Hello @GGraziadei,
thank you for the pieces of advice.

1. I had overlooked the case, but I have now checked. add_common_options (bin/storm.py:385-389) declares main_args with nargs='*' and no default=, so it never holds synthetic values from argparse. Only local() (bin/storm.py:1065) prepend a non-CLI value in main_args and that runs after raw_args.func(raw_args) (and consequently the introduced method). No subcommand relies on a pre-merge injected default, so the replace semantics are safe;
2. The proposed solution manages to be more elegant, concise and complete than the original one, implemented;
3. It now falls back to the original functioning as this behaviour was accepted before;
4. A new docstring with more information regarding input and output has been added. Consider accepting or rejecting it based on verbosity.

Also, I have renamed the method as I have seen that all the other ones in the file use only present simple and not present continuous

[GGraziadei]

Hi @sercuzz8, thanks for addressing these points. Good first-time contribution!
LGTM