`tm_g_distribution` always starts with validation error

[chlebowa]

What happened?

This is not necessarily a bug but is does interfere in some cases.

tm_g_distribution starts with Tests* (id = 'teal-main_ui-root-dist_1-module-dist_tests_input') unselected, which produces a validation error in the Tests output. This results in a validation error. Since it is impossible to set the theoretical distribution in the module call, this is inevitable.

Would it be possible to either
a) blocking the Tests output be achieved without raising a validation error,
b) add an argument to tm_g_distribution?

(*) probably should be "Test", by the way

sessionInfo()

No response

Relevant log output

No response

Code of Conduct

• I agree to follow this project's Code of Conduct.

Contribution Guidelines

• I agree to follow this project's Contribution Guidelines.

Security Policy

• I agree to follow this project's Security Policy.

[donyunardi]

Here's what it looks like

a) blocking the Tests output be achieved without raising a validation error,

If we hide the output until a test is selected, it might obscure the test feature for this module. I think the intention is to encourage users to select a test or to make user aware of the test output/selection.

We could also replace the validation with something more informative to highlight the presence of the "Test" feature if we have to remove the output when no test is selected.

b) add an argument to tm_g_distribution?

Possibly, it would have to be a required argument with a default value from one of the selection here to avoid the validation to trigger when nothing is selected.

teal.modules.general/R/tm_g_distribution.R

Lines 360 to 378 in 9d193d3

	teal.widgets::panel_item(
	"Tests",
	teal.widgets::optionalSelectInput(
	ns("dist_tests"),
	"Tests:",
	choices = c(
	"Shapiro-Wilk",
	if (!is.null(args$strata_var)) "t-test (two-samples, not paired)",
	if (!is.null(args$strata_var)) "one-way ANOVA",
	if (!is.null(args$strata_var)) "Fligner-Killeen",
	if (!is.null(args$strata_var)) "F-test",
	"Kolmogorov-Smirnov (one-sample)",
	"Anderson-Darling (one-sample)",
	"Cramer-von Mises (one-sample)",
	if (!is.null(args$strata_var)) "Kolmogorov-Smirnov (two-samples)"
	),
	selected = NULL
	)
	),

We could add a default value to the selected argument in the server logic since the choices are hard-coded anyway but this approach might feel too abrupt.

[chlebowa]

a) blocking the Tests output be achieved without raising a validation error,

If we hide the output until a test is selected, it might obscure the test feature for this module. I think the intention is to encourage users to select a test or to make user aware of the test output/selection.

We could also replace the validation with something more informative to highlight the presence of the "Test" feature if we have to remove the output when no test is selected.

That's what I meant. The validation message suppresses whatever output would be generated with no selection (probably errors). If the suppression is performed without validation errors, that's fine by me.
It is also les sinvasive than opening a new argument.

Would you like me to propose a change?

[donyunardi]

@kumamiao can I also hear your thought on this before we move forward?

[kumamiao]

I agree that the intention is to instruct the users to utilize the test feature, so completely blocking the test output by default is not ideal here. Since we do not provide arguments for other settings (i.e., distribution), I would suggest to go with below:

We could also replace the validation with something more informative to highlight the presence of the "Test" feature if we have to remove the output when no test is selected.

[chlebowa]

Thanks @kumamiao, I will prepare a fix.