Drop iterate/@precedence

[clange]

Another issue that was neglected on the mailing list… I will turn all of these into tickets, as there hasn't been any objection :->

I stumbled upon a little discrepancy between specification (as in the MCS paper) and implementation. In the specification, we have the abstract syntax

for (listjoker, increment, the-separator) { loop-body }

which translates to (via the E encoding function)

<iterate name="listjoker" step="increment">
 <separator>
   E(the-separator)
 </separator>
 E(loop-body)
</iterate>

However, in the actual *.ntn files, I frequently see

<iterate name="listjoker" precedence="p">

Now we could extend the abstract syntax to

for (listjoker^precedence, ...)

to cover this, but is it necessary? I suppose that

<iterate name="listjoker" precedence="p">
 <render name="elementjoker"/>
</iterate>

is equivalent to

<iterate name="listjoker">
 <render name="elementjoker" precedence="p"/>
</iterate>

If that is the case, having a precedence attribute on both iterate and render would actually be an ambiguity with an undefined resolution.

Can you imagine any case where it does make sense to have iterate/@precedence? If not, let me suggest

• to drop it from the implementation
• and from all existing notation definitions (and move the predecence attribute down to the render child elements)

[clange]

As nobody has objected so far, let's do it.