Changed type of select

[xml-project]

I tried to spell out @gimsieke 's idea: Is it as easy as this?

[ndw]

I think that covers it.

[ndw]

Except. Sigh. Doing things quickly always takes longer.

What is this supposed to produce?

<p:identity><p:with-input><doc/></p:with-input></p:identity>
<p:message select="."/>

If select is an ordinary option, that's the literal string ..

So then you think, I'll do this:

<p:identity><p:with-input><doc/></p:with-input></p:identity>
<p:message select="{.}"/>

That produces "" (the empty string, because AVTs are always the string value).

I think the expectation is more likely to be that select is a simple xs:string and it's treated as an expression that is evaluated to get the result.

[xml-project]

@ndw I think the correct way to do it in this case is:
<p:identity><p:with-input><doc/></p:with-input></p:identity> <p:message> <p:with-option name="select" select="." /> </p:message>
You can't use an AVT in this case.
Does this make sense?

[ndw]

Yes. I think you're right.

I was going to argue that the name select and the parallel with xsl:message would strongly encourage people to believe that it should just be an expression and we should do it that way. Then I remembered that the static context for a step doesn't include the in-scope variables.

If we make it an expression, you can type select="'test: ', .", but you can't type select="'test: ', $varname". The work around for . is just use p:with-option, the work around for variable names is really ugly expression like

select="'test: "' || $varname || '"'"

Where you have to make sure that the variables are resolved while the string is being constructed but the resulting string is still a valid expression. Yuck.

[xml-project]

@ndw So your idea is to go back to select as xs:string and wait for better ideas to come up?

[ndw]

No, I think item()* is fine, it means I can put more complicated things in the output if I use p:with-option. I'm just abandoning my thought that it should be an expression resolved by the step.

[xml-project]

Oh, now I understand what you had in mind. Haven't thought of it this way. Is this why it reads:

If the test attribute is “true”, the select expression is evaluated and made available.

Should we change this?

[ndw]

I don't know. The [p:]message attribute speaks of computing the value, but it has special rules about expand-text.

Does this seem like an improvement?

If the test attribute is “true”, the value of the select option is made available.

[gimsieke]

the result of evaluating the select option is serialized and made available