Multiple short options not working anymore

[bjorntj]

I have a method with the following:

@ShellOption(valueProvider = ServerValueProvider.class, help = "Choose which server to connect to", value = "-s") @NotBlank String server, @ShellOption(help = "Comma separated list of filegroups to update, no spaces", value = "-f") @NotBlank String filegroups, @ShellOption(help = "Add parameter to commit the update", value = "-c") boolean commit

If I run the app, I get the following error:

install-vareviews -s xlbygg -f j41
2000E:(pos 0): Missing mandatory option, longnames='', shortnames='f'

And if I switch the arguments, I get the following:

install-vareviews -f j14 -s xlbygg
2000E:(pos 0): Missing mandatory option, longnames='', shortnames='s'

So it seems like it does not catch the second argument?

This works fine using Spring Shell 3.0.x but not using version 3.1.0.

[tincore]

There is also the case of using short names and failing to calculate the arity

argument input, i with arity 1

commandX -i arg0 -o arg1

Fails reporting that -i only supports one argument (thinking that -o arg1 are more arguments)

commandX -i arg0 --output arg1

Succeeds.

It has to be something recent. In the project I'm working atm I've been jumping from 3.0.x to 3.1.0-RCx a few times during last weeks and all my automated tests passed without changes (before 3.1.0)

[jvalkeal]

Thanks reporting this. I'll post more info when I find what happened as I see the same issue.

[jvalkeal]

I've pushed a commit which hopefully fixes this issue. I'll keep issue open if you want to try snapshots and add a comment.

This bug was in a parser tokenisation which i.e. with install-vareviews -s xlbygg -f j41 created tokens of

TokenType.COMMAND TokenType.OPTION TokenType.ARGUMENT TokenType.ARGUMENT TokenType.ARGUMENT

while it should have been:

TokenType.COMMAND TokenType.OPTION TokenType.ARGUMENT TokenType.OPTION TokenType.ARGUMENT

then a parser got confused.

[tincore]

I can confirm that with latest snapshot all my tests pass. In my case I can go from 3.0.x to 3.1.x without code changes.

[tincore]

Still there are other related errors!

:>get-value -k MY_KEY
2004E:(pos 0): Too many arguments for option 'key', requires at most '1'

:>get-value --key MY_KEY

correct response

[jvalkeal]

@tincore I can't reproduce. Can you show a bit more how that command is defined.

[tincore]

Sure. Maybe its the position argument or that redundancy with arity and required. This is one of the cases were it would be convenient to be able to autocomplete based on position to avoid typing the option name.

 .command("get-value")
            .withAlias()
            .command("g")
            .and()
            .withOption()
            .longNames("key")
            .shortNames('k')
            .label("PROPERTY_KEY")
            .description("Properties key. Possible values " + Arrays.stream(PropertiesGetKey.values()).map(Enum::toString).toList())
            .type(PropertiesGetKey.class)  // Its an enum
            .required()
            .arity(CommandRegistration.OptionArity.EXACTLY_ONE)
            .position(0)
            .completion(enumValueProvider::complete)
            .and()
            .withOption()
            .longNames(OPTION_VERBOSE)
            .shortNames('v')
            .label("BOOLEAN")
            .description("Verbose output")
            .type(Boolean.class)
            .and()
        

[jvalkeal]

Not sure as copying that format doesn't make things to fail. Would you be able to make a sample I could run?

[tincore]

I was building the sample and I upgraded again to latest 3.1.1-SNAPSHOT and now it seems to work fine! Both the sample and the original code that was failing before.

Thanks for all! :)

Btw. If you have a chance. Check that PR that I left some time ago ;) It makes for better auto-completion for filesystem paths.