\tl_case:nn, \str_case:Nn, and \str_if_empty:n

[lvjr]

Recently I compared tl and str packages:

• there is \tl_case:Nn but not \str_case:Nn
• there is \str_case:nn but not \tl_case:nn
• there is \tl_if_empty:n but not \str_if_empty:n

Could these missing functions be added to expl3?

[eg9]

There is \str_case:Vn which serves the same purpose as the proposed \str_case:Nn.

There is also an inherent difference between \str_case:Vn and \tl_case:Nn, as the latter compares token list variables (with \ifx) and not “explicit” token lists, so \tl_case:nn would be “strange”. It could be implemented by setting an internal variable to the given token list and then calling \tl_case:Nn for this variable, but I'm not sure this is something we should add.

Indeed, I find strange that \str_if_empty:n(TF) is missing.

[josephwright]

I guess \str_if_empty:n was missed as it doesn't matter if one uses a tl function here - the nature of the non-space tokens is unimportant. However, it's easy enough to add.

As @eg9 says, \str_case:Nn is the same as \str_case:Vn: again easy to add and again I'm happy to.

I'm not keen on \tl_case:nn for the reasons outlined.

[lvjr]

I think consistency is important. otherwise it will confuse users. In my opinion, it would be better to use \str_if_eq:NN inside \str_case:Nn function just like in tl package.

\str_set:Nn \l_tmpa_str {\abc}
\str_set:Nn \l_tmpb_str {\abc}
\str_case:VnF \l_tmpa_str
  {
    \uvw        { UVW }
    \l_tmpb_str { TMP }
    \abc        { ABC }
  }
  { No Match }

The above code gives ABC, which shows that \str_case:Vn in not \str_case:Nn.

[blefloch]

I think \tl_case:Nn should be removed or renamed \cs_case_meaning:Nn. It is a leftover from the old days where one had to compare macros with stored values to get any reasonable speed.

[josephwright]

I think @blefloch is right here: the approach of comparing token lists is actually not that useful in the main in my experience. We can't remove the name, but we can deprecate.

[josephwright]

I think consistency is important. otherwise it will confuse users. In my opinion, it would be better to use \str_if_eq:NN inside \str_case:Nn function just like in tl package.

\str_set:Nn \l_tmpa_str {\abc}
\str_set:Nn \l_tmpb_str {\abc}
\str_case:VnF \l_tmpa_str
  {
    \uvw        { UVW }
    \l_tmpb_str { TMP }
    \abc        { ABC }
  }
  { No Match }

The above code gives ABC, which shows that \str_case:Vn in not \str_case:Nn.

I'm not sure what you mean here: the defined behaviour is that \str_case:Nn compares the value of each entry to a string. You've omitted the braces in your list of possible strings, but we do not expect to compare with the value of \l_tmpb_str, rather those tokens as a string.

[lvjr]

I'm not sure what you mean here: the defined behaviour is that \str_case:Nn compares the value of each entry to a string. You've omitted the braces in your list of possible strings, but we do not expect to compare with the value of \l_tmpb_str, rather those tokens as a string.

Because \tl_case:Nn is a shorthand command for some nesting \tl_if_eq:NN, therefore in my mind \str_case:Nn should be a shorthand command for some nesting \str_if_eq:NN. That is to say, in my mind the syntax of \str_case:Nn is

\str_case:Nn <test str variable> {
  <str variable case 1> {<code case 1>}
  <str variable case 2> {<code case 2>}
  <str variable case 3> {<code case 3>}
}

[blefloch]

We concluded \tl_case:Nn should be deprecated and renamed \cs_case_meaning:Nn, but there is already \token_case_meaning:Nn, which seems to fit the bill. Relatedly, \tl_if_eq:NN(TF), \cs_if_eq:NN(TF) and \token_if_eq_meaning:NN(TF) have identical definitions, and probably some should be defined as copies of others with the appropriate prg_new_eq function.

[josephwright]

We concluded \tl_case:Nn should be deprecated and renamed \cs_case_meaning:Nn, but there is already \token_case_meaning:Nn, which seems to fit the bill. Relatedly, \tl_if_eq:NN(TF), \cs_if_eq:NN(TF) and \token_if_eq_meaning:NN(TF) have identical definitions, and probably some should be defined as copies of others with the appropriate prg_new_eq function.

Multiple names are fine as long as they help with understanding: I think \tl_if_eq:NN and \cs_if_eq:NN certainly fit that bill. I'll look at the tidy-ups shortly.

[josephwright]

We still need some action here?

[muzimuzhi]

I guess \str_if_empty:n was missed as it doesn't matter if one uses a tl function here - the nature of the non-space tokens is unimportant. However, it's easy enough to add.

As @eg9 says, \str_case:Nn is the same as \str_case:Vn: again easy to add and again I'm happy to.

_{quoted from #1071 (comment)}

Still planning to add \str_if_empty:nTF and \str_case:Vn?

FYI: They were added in 025bde6 and 600f6c3, respectively.

[josephwright]

I guess \str_if_empty:n was missed as it doesn't matter if one uses a tl function here - the nature of the non-space tokens is unimportant. However, it's easy enough to add.
As @eg9 says, \str_case:Nn is the same as \str_case:Vn: again easy to add and again I'm happy to.
quoted from #1071 (comment)

Still planning to add \str_if_empty:nTF and \str_case:Vn?

They were done the same day they were discussed here :)