[sword-devel] modules to relational database (Ruby support)

[Bill Burton]

Hello,

I see.  You want the module information in a database because it's then
trivial to program a web application using Ruby on Rails ActiveRecord object
relational mapping support.  RoR also supports other logic in controllers
such as web services clients so it's entirely possible to write an
application that calls Sword without using a database but the scaffolding
code generator won't be able to help much if at all.

It just so happens I've been working on a Ruby interface to Sword with a
longer term goal of writing a Ruby on Rails application to access it.
Initially, I couldn't figure out how to build anything with SWIG--there's no
README in the bindings/swig directory.  Then it happened I discovered Ruby
can load shared libraries and DLL's and dynamically bind to them on the fly
using the dynamic loader (DL) module.  So I started to write a Sword
interface to the C flatapi in bindings/flatapi.cpp.  In the process, I found
a few minor bugs with the flatapi but have created patches to fix them
(which I'll submit soon).

Using this Ruby to Sword flatapi interface you can output verses as follows
(I'm doing this from memory so it's not exact):

include Sword
# SWMgr_new(FMT_HTML)
Manager.new(Sword::FMT_HTML) do |mgr|
module = mgr.get_module("KJV")
# iteration using Sword API's under the covers to verseKey and renderText
module.each_verse_render("Psalm 133:1-3") do |verse, text|
puts "#{verse} #{text}"  # verse reference and text
end
end  # performs an implicit SWMgr_delete

So far, I've implemented about two-thirds of the Sword C flatapi.  However,
the one issue I'm just starting to address is how to call these API's within
a multi-threaded multi-user environment such as a web application.  Because
the SWMgr_new/newEx and SWModule_* functions have state, it looks like a new
Manager object will have to be created per user which means establishing a
session and then saving the Manager instance in the session.  So I have to
refactor the code to allow multiple instances of a Manager class without
loading the shared library every time.

But then last night I was about to send an email to this list asking how to
build the SWIG interface but I looked at it one more time and discovered how
to do it. With some more investigation I was then able to generate SWIG
bindings for Ruby and build the interface to Sword.  So far I've been able
to access some of the methods from the SWMgr and SWModule classes but can't
output a verse because the binding of set_key to SWModule.SetKey is
incorrect.  This may be a bug in the way SWIG generated the binding but I
don't know yet.

The nice thing about the Ruby dynamic loader interface to Sword is there's
nothing extra to build which makes it much easier to install as compared
with the SWIG version.  However, the C flatapi on which it's based limits
one to getting and setting global options, iterating over modules and
rendering verses.  The SWIG interface to the C++ API's is much more complete
but has to be built. I don't know yet if the Sword C++ API's also have state
which is important to know for a web application.

Right now, I'm developing on RedHat ES 3 but plan to test the DL version on
Windows against an installed BibleCS.  My time is limited but I'll try to
get this binding in reasonable shape soon so it can be used with Ruby on
Rails.  Building the SWIG version on Windows is not an option for me at this
time due to lack of tools and a dead machine.

God is merciful,
-Bill

On 11/7/06, lumin8 <lumi.n8 at gmail.com> wrote:
>
> I'll leave the question alone as to the value of a relational database
> > for this data over using the SWORD API.
>
>
> Actually, I am interested in this question if you have time / desire to
> enlighten me.  I think I can manage the chunk of code you gave me (thanks),
> but I have never compiled a program in my life.  I have been building web
> applications for about 7 years with PHP, and now Ruby for the last year and
> a half.
>
> I am willing to learn the Sword API if I need to, but I can build web
> applications very quickly with Ruby on Rails.   The rest of my data (user
> info, notes, user generated translations, a wiki like interface for
> commentary) will all be in a relational database.  Would there be speed
> benefits from using the Sword API over the indexing provided by mysql or
> postgre?  What about application development time?  In the Rails framework,
> I hardly even have to write SQL, if I used the Sword API I would have to
> learn a bit about c++ and bindings to Ruby...  I am willing, but what is the
> advantage?
>
> By the way, I have tried and use many of the Sword front ends and highly
> respect the work you all are doing.
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://www.crosswire.org/pipermail/sword-devel/attachments/20061107/4c9278d4/attachment-0001.html