SIL Not Recommending SILDoulos IPA93 Now (fwd)

[Koontz John E]

On Thu, 1 Jun 2006, Koontz John E wrote:
> I remember a number of rule-based conversion tools apart from Consistent
> Changes.

The program I couldn't remember was called bitrans.

My usual approach in scripts is something like:

for each input line

case for lines of one kind
case for lines of another kind
etc.

In the cases one either writes out the line as is or translates it,
depending on what kind it is.

If it requires translation that code (not terribly efficient) is something
like:

while we can match the pattern all-possible-input-sequences
  take the matched character
  replace it with output-sequence[match]
  put that replacement at the end of what we've translated for this line
   so far

When we're done, write out the translated line. This uses an associative
array e.g., if we're converting Dorsey to Siouanist out-sequence["p"] is
predefined as "pp" (usually the right answer), while if we are replacing
Siouanist with Omaha popular we have out-sequence["pp"] predefined as "p".
To handle p after s as just p we either have to define a rule for sp or we
need some special cases in the matching code.

The pattern all-possible-input-sequences matches the list of things that
are indexes in out-sequence, e.g, it would be "p" or "b" or "pp" ...
The details depend on the language in which the script is written and the
mapping being performed.