Libelektra: kdb set user/ and -N user should behave the same, -f for old behavior

[markus2330]

@sanssecours Can you maybe help define the behavior here?

[sanssecours]

If you mean that something like

kdb get user/tests/get

should behave the same as

kdb get -N user /tests/get

, then I think that sounds reasonable. What exactly is the old behavior? As far as I can tell kdb get does not support an argument -N yet:

kdb 2>&1 get -N
#> kdb: invalid option -- N
#> Sorry, I could not process the given options (see errors above)
#> 
#> Usage: kdb get <name>
#> 
#> Get the value of an individual key.
#> When the key starts with / a cascading lookup will be done.
#> 
#> Example:
#> 
#>    kdb get system/elektra/version/constants/KDB_VERSION
#> 
#> -a --all                 Consider all of the keys.
#> -H --help                Show the man page.
#> -n --no-newline          Suppress the newline at the end of the output.
#> -p --profile <name>      Use a different profile for kdb configuration.
#> -v --verbose             Explain what is happening.
#> -V --version             Print version info.
#> -C --color <when>       Print never/auto(default)/always colored output.

.

[markus2330]

If you mean that something like

exactly, but I meant kdb set

What exactly is the old behavior?

That one way supports the validation and the other does not.

[sanssecours]

That one way supports the validation and the other does not.

Yes, it make sense that in the following situation:

kdb mount tutorial.dump /tests/spec dump validation
kdb setmeta spec/tests/spec/test check/validation '[1-9][0-9]*'
kdb setmeta spec/tests/spec/test check/validation/match LINE
kdb setmeta spec/tests/spec/test check/validation/message 'Not a number'

, the command

kdb set user/tests/spec/test 'incorrect value'

fails, just as

kdb set -N user /tests/spec/test 'incorrect value'

and

kdb set /tests/spec/test 'incorrect value'

do. Using the option -f (force) to change the value, even if the validation fails sounds like a good choice.

[markus2330]

Exactly! Thank you for the description of the behavior. Maybe we should also add -N to kdb get so that the same syntax can be used for every kdb command.

[markus2330]

Maybe we can fix this together with #2724. Then scripts do not need to be adapted twice.

[kodebach]

IMO we should always use a cascading parent key. If the given key has a namespace, strip it for creating the parent key, but use it during ksAppendKey (in kdb set) and ksLookup (in kdb get) et al. I think this is the only good solution, since the spec plugin, which is more or less essential for other plugins to work, only works for cascading parent keys.

To disable this behaviour and use the namespace in the parent key as well, we would simply provide a command line option.

[markus2330]

IMO we should always use a cascading parent key.

Yes, I agree. It gives a much more consistent user experience.

If the given key has a namespace, [...]

Yes, makes sense. But let us implement this after #3115, if we feel it is needed.

[stale[bot]]

I mark this issue stale as it did not have any activity for one year. I'll close it in two weeks if no further activity occurs. If you want it to be alive again, ping the issue by writing a message here or create a new issue with the remainder of this issue.
Thank you for your contributions :sparkling_heart: