Re: Computer Language Translation

[joevl@mojsys.com (Joe Vlietstra)]

In article <791@nfsun.UUCP> tkueny@nfsun.UUCP (Todd Kueny) writes:
>[...Discussion about language translation products...]
>I would like to hear any comments concerning translation that people
>may have.  Only serious comments please, no flames about why its a
>terrible thing to do; some of us have no choice.
>
>########################################################################
>Todd R. Kueny			{ My opinions are my own. }
>nfsun!tkueny@uunet.uu.net	105 Bevington Road
>(412) 243-1630			Pittsburgh, PA  15221

We are looking at developing a QuickBASIC-to-Ada translator for
internal use.  (Don't ask me why, it's a long story.)  We have also
used a lot of PD and commercial translators.  Our experience corresponds
with your observation -- there's a lot more to translation than simple
re-expression in another language.

I can summarize a translation task into the following steps:

    -- INITIALIZATION
       Experiment with the translator.
       Use lint (or flint) to identify problems in original code.
       Use cross-ref to develop data base for global variables.
       This step represents roughly 10% of the total effort.

    -- PRE-TRANSLATION
       Place original version into configuration management.
       Fix minor problems reported by lint (e.g. unused vars, etc.)
       Try to clean up spaghetti code.
       Perform regression test against original version.
       Approximately 25% of the total effort.

    -- TRANSLATION
       Run translator.  Fix translator artifacts.
       Perform another regression test.
       Only 5% of the total effort.

    -- POST-TRANSLATION
       Fix major problems reported by lint.
       Incorporate Ada structured programming features.
       More regression testing.
       60% of the total effort!

Even if the translator is able to correctly translate 100% of the
original code, there will be a significant post-translation effort.
The post-translation effort stems from the fact software translated
into Ada is never as well-structured as software originally written
in Ada.  (In fact, translated software is pretty darn ugly.)
A significant clean-up effort is required before the translated
software is presentable.  The management challenge is to prevent
the clean-up effort from becoming a re-write effort.
				Joe Vlietstra

-- 
------------------------------------------------------------------
Joe Vlietstra        | Checked Daily:  ...!uunet!mojsys!joevl
Mojave Systems Corp  | Checked Weekly: mojave@hmcvax.claremont.edu
1254 Harvard Avenue  | Iffy Routing:   joevl@mojsys.com