[sword-devel] load v11n from file

[Konstantin Maslyuk]

> You make a great and convincing case for proceeding down path #2.
>>     2. each module may have its own v11n, no dynamic shared v11ns
>
> Path #2, though, disregards our objective to ultimately provide a way to
> position Bibles of differing v11ns to the same content.

> I would love to forget about trying to enumerate and define
> versification systems (I'm sure Chris would as well), and to stop asking
> each Bible module maker to pick which system best represents their data.
>
> We currently have 13 systems in our engine.  Eventually we need to
> extend VerseMgr with a facility to map these between one another
> (possibly your implementation).
>
> You have stated that you have tried with your code
>
>> KJV(A) <-> Synodal <-> Vulg <-> NRSV v11ns
>
> This implies there exists such a thing as a named list of versification
> system in our engine (i.e., NOT path #2).

Of cource, most common v11ns must be hard coded! I can not understand here  
why
path #2 can not have mixed v11ns: several built-in and several  
module-supplied.

Except without remark that module name with module-supplied v11n can not  
be same
as built-in v11n.

This would be problem of bad statement of thoughts, i didn't meant that  
all modules
should have module-supplied v11n. Sorry.


I do not see a problem in making mapping through v11n other than KJV, but  
verse
mapping must always be thought built-in v11n. One v11n for mapping would be
enought, or one meta-v11n.

Though i do not know yet all problem case with v11n mappings, and i know  
that there
are questionable parts of the Scripture, where in one source one chapter  
should have
one content and second source under the same chapter have different  
content. But in
Sword such parts should be known under unified name (even like v1:Esd.34 =
metav11n:Esd.110, v2:Esd.34 = metav11n:Esd.34).

> If you allow a module developer to bypass naming which versification
> system their module uses, you would still somehow like them to define
> how it maps to other Bibles.  I don't see how this is practical without
> a named set of known v11n system in the engine.
>
> I cannot imagine a module developer ALWAYS defining how their module
> maps to every other system.

Of course if module maker do not use built-in v11n, he should use  
module-supplied.
In any case #1 or #2 he can take ready binary v11n from repository of  
av11s (that
we must provide in any case) and put within module. Or he can take source  
file for
v11n, edit and convert to binary format, and use this binary file as many  
times as
needed. Of course for engine it will be different v11ns, but this is a pay  
for order.

It is also advantage of external v11ns: we can create and test v11ns  
separate from
engine, and if v11n is popular enough or well tested it can be made  
built-in, and
vice versa.

Source for v11n would be just an OSIS file (of course, i'm not sure that  
this is correct):
<div type="book" osisID="Gen">
	<chapter osisID="Gen.1">
		<verse osisID="Gen.1.1"/>
		<verse osisID="Gen.1.2"><reference osisRef="Gen.1.2-Gen.1.3"/></verse>
		<verse osisID="Gen.1.3"><reference osisRef="Gen.1.4"/></verse>
	</chapter>
</div>

Or add method VerseMgr::saveVersification(const char *file), that will  
save static
arrays of v11n data in to file. So, module maker should write header file  
with v11n
compile and test it, and then call saveVersification, i would agree it is  
not trivial.


> I can imagine a module maker saying, "I am basically "Synodal", but have
> these 5 exceptions, and here is how these 5 exceptions should be handled
> in relation to "Synodal."
>
> And then our future mapping system can do it's best with this  
> information.

My mapping system would be flexible with preservation of back  
compatibility.
It is a list of data (rules)
1 1 1 0 1 1 1 2
1 1 2 0 1 1 2 3
book/chapter/verse/end_verse/...(same to target)

on initialization VerseMgr parses whole data and remembers pointers to the  
first
entry of rule, i would use here magic numbers (245-255) to reserve any  
kind of
specific rules, old version of engine would throw away such rules, because  
them
are not supported.

BTW i want to discuss base size for mappings: char, short or int. With  
char mappings
for Synodal it is 1,3 kb, but it is limited up to 255 books/chapter/verses.

> But I believe we still need to enumerate a list of officially supported
> v11n systems and have module developers choose which one to use.

I do not urge to run and make new system for external v11n. Everything  
should be well
discussed and accepted.

But for now i see that possibility to load v11n from file would be useful,  
in process of
moving to conception of multiple v11ns. And it is not hard to implement  
(one function
to load v11n from file and add load logic for SWMgr). We do not need tools  
now.

> I'm not happy about the extra work, but do you agree it is necessary?
Please, precise what kind of work is necessary now?