transpose empty elements that span to anchors

[gerritbruening]

When I have something like

                <transpose f:revType="soon-or-late">
                    <ptr target="#l7172-l7174"/>
                    <ptr target="#l7175-l7177"/>
                </transpose>

and

                                    <milestone xml:id="l7175-l7177" spanTo="#anchor_l7175-l7177" unit="group"/>
                                    <l n="7175">Das Trallern iſt bey mir verloren</l>
                                    <l n="7176">Es krabbelt wohl mir um die Ohren</l>
                                    <l n="7177">Allein zum Herzen dringt es nicht;</l>
                                    <anchor xml:id="anchor_l7175-l7177"/>
                                    <milestone xml:id="l7172-l7174" spanTo="#anchor_l7172-l7174" unit="group"/>
                                    <l n="7172">Das ſind die ſaubern Neuigkeiten</l>
                                    <l n="7173">Wo aus der Kehle, von den Saiten,</l>
                                    <l n="7174">Ein Ton ſich um den andern flicht.</l>
                                    <anchor xml:id="anchor_l7172-l7174"/>

the transposition for last stage doesn't work. @thvitt, can you find out how often we have transpose that points to a milestone?

[gerritbruening]

Oh, pointer point to anchor, that's bad.

[thvitt]

Transposed empty elements

Sigil with Transposition	ID of transposed el.	transposed element	non-empty partners	spanTo	common parent
2 I H.56	#Einweisungsstelle	anchor	0
2 II H.22	#start2	anchor	1
2 II H.22	#start1	anchor	1
2 V H.37c	#l12005-12012	anchor	1
2 V H.37c	#l12013-12019	anchor	1
2 III H.2:1	#gräslich	seg	0
2 III H.2:1	#ging-die-Sonne	anchor	0
2 III H.2:1	#die-er	anchor	0
2 III H.2a:5	#Versumbruch	anchor	0
2 III H.7:1	#Versumbruch	anchor	0
2 II H.74	#l7172-l7174	anchor	1
2 II H.74	#l7175-l7177	anchor	1
2 V H.1	#after_11283	anchor	1
2 V H.1	#before_11286_b-11287	addSpan	1	anchor	sp
2 II H.5a	#p132_Indessen-So	milestone	1	anchor	div
2 II H.5a	#p132_Sie-So	milestone	1	anchor	div
2 I H.52	#Zielposition	anchor	0
2 H	#before_7495_c	anchor	0
H P124	#start_transposition_1	anchor	1
H P124	#start_transposition_2	anchor	1

for $ptr in $faust//transpose/ptr
let $target := id(substring-after($ptr/@target, '#'), $ptr)
for $milestone in $target[not(child::node())]
let $span-end := id(substring-after($milestone/@spanTo, '#'), $milestone),
    $parent := $span-end/..,
    $is-common := if ($span-end) then if ($parent is $span-end/..) then name($parent) else "✘" else "",
    $sigil := data(root($milestone)//idno[@type='faustedition']),
    $partners := for $partner in ($ptr/../ptr except $ptr) return id(substring-after($partner/@target, '#'), $ptr)
return <h:tr>
    <h:td>{$sigil}</h:td>
    <h:td>#{data($milestone/@xml:id)}</h:td>
    <h:td>{name($milestone)}</h:td>
    <h:td>{count($partners[not(child::node())])}</h:td>
    <h:td>{name($span-end)}</h:td>
    <h:td>{$is-common}</h:td>
    </h:tr> 

[thvitt]

The transposition doesn’t work because this use case has never been specified and thus never been implemented. The phrase-level transposition in the test manuscript has its xml:ids on lines.

I would very strongly suggest to check whether this couldn’t be handled in some different way (e.g., pointer to l or div or lg or ab). I haven’t looked at the actual cases yet, but transpositions across hierarchy borders might contain all kinds of ambiguities and will not be easy to implement.

[gerritbruening]

Same applies to addSpan (https://github.com/faustedition/faust-xml/issues/134).

[thvitt]

This table lists all empty elements that are referred by a transposition pointer. This points to different cases, at least the following can be distinguished:

1. anchor with ≥ 1 non-empty partner(s): The non-empty segment is moved to the anchor position. Should work ok, although there might be encoding glitches that lead to unexpected whitespace behaviour. Cf. Was Raub, compare the inline apparatus.
2. anchor with 0 non-empty partner(s): semantics unknown, @gerritbruening you’ll need to check what happens there
3. milestone/spanTo with common parent: Can be unproblematically encoded using an enclosing element (seg / ab) instead of the milestone approach, which will make the existing implementation work. If this isn’t acceptable implement a preprocessing step

[gerritbruening]

Thanks for the query and its analysis! I will check nr. 2 for 1.2. As for nr. 3, I think we already looked out for a suitable enclosing element in @wissenbach|s time, but we didn't find one: seg is not allowed where we would need it, as it's phrase-level. ab is, and l has some time been allowed to appear in ab. It would be slightly abusive, though, since ab is not supposed to be a mere container element (i.e., is not suppused to be what we need).