tpl: adds truth and bool template functions

[khayyamsaleem]

The behavior of truth and bool is described in the corresponding test cases and examples. The decision-making around the behavior is a based on combination of the existing behavior of strconv.ParseBool in go and the MDN definition of "truthy" as JavaScript has the most interop with the Hugo ecosystem.

Fixes #9160 and (indirectly) #5792

[khayyamsaleem]

@bep open to discussion on the behavior / naming for sure; thought this would open a more productive discussion on #9160 as it was low-hanging fruit that has been open for a long time.

[khayyamsaleem]

For truth, could also just use the rules in hreflect.IsTruthful, but thought that the MDN standards may be more familiar to end-users. I guess it's unlikely to come up with something that EVERYONE likes here, so best to provide some capability and gauge the response from discourse or GH issues after the release.

[bep]

@khayyamsaleem thanks for this. My 50 cents about this, and please don't adjust your code before we agree on something is.

As to truth, these test cases needs to pass and behave the same for all "truthy values" on the left side:

{{ if "foo" }}OK{{ else }}Failed{{ end }}
{{ if ("foo" | truth) }}OK{{ else }}Failed{{ end }}
{{ if ("foo" | not | not ) }}OK{{ else }}Failed{{ end }}

Which means that if would be a thin wrapper around IsTruthFul.

As to bool, I'm tempted to just use strconv.ParseBool. Combining it with some JavaScript behaviour is surely going to get us into the Yaml trap (YAML < 3 had "yes/no" as bools).

//cc @jmooring

[jmooring]

A colleague of mine told me about a 700+ line YAML file they had to debug a few years ago. It had some country codes. Norway was falsy.

[khayyamsaleem]

@bep I'm fine with the suggestions; better to have one definition of truthiness for the codebase.

[bep]

@khayyamsaleem OK, if you adjust this PR to that much simpler scenario, I will merge.

To be clear:

• bool == https://pkg.go.dev/strconv#ParseBool
• truth = if reflect.IsTruthful(v) true else false

[bep]

OK, I changed my mind again. Looking at the other cast.* functions we have, and also on other places we do similar, I think it's a pretty clear precedence to just use cast.ToBoolE, so I suggest we just do that for the bool func.

[stereobooster]

HI @bep. Does this PR needs more work? As far as I understood ToBool function is ok. Do we need to adjust ToTruth function (if reflect.IsTruthful(v) true else false)? Can I help here?

[khayyamsaleem]

@bep pardon the delay here. Updated PR per your last review.

[jmooring]

This is not what we agreed to:
#10666 (comment)

cast.ToBool behaves as I expect, though I think we should return false on error.

cast.ToTruth should use hreflect.IsTruthful. The string "false" is truthy.

[khayyamsaleem]

Ahhh pardon, you're right. It did seem awkward for Truth and bool to disagree on "false" though.

[jmooring]

@bep I have not tested this yet, but I think the truth function belongs in the compare namespace as IsTruthy. This function does not cast a value into another data type; it performs a comparison.

{{ compare.IsTruthy $foo }}
{{ isTruthy $foo }} (alias)

and

{{ cast.ToBool $foo}}
{{ bool $foo }} (alias)

Of course, that also raises the question, do we want a compare.IsFalsy too? It depends on whether you like the symmetry, or not.

[khayyamsaleem]

Are we waiting for @bep 's feedback here, or would we prefer to move the function to the compare namespace?