#!/bin/bash

# -*- mode: sh; coding: utf-8 -*-

# $Date: 2008/03/01 06:26:19 $

# Do the matching for Latin alphabet (English), but in the UTF-8 encoding (for wide characters).

LC_ALL=POSIX.UTF8 sed -e "# I use the doublequote because I use ' in the scripts.

# NB: take care when writing this map that it should basically be injective

# (except for unimportant variants like s^' and s'^): if it's not, it means

# there is a mess in your transcription system, and you should fix your trancripts.

# FIXME: Terrible bug: now, '\Fut' gives '\wˣut'. This is not elegantly

# avoidable in this sed script, so I'll rewrite the program in another language.

#############################################

# Very special combinations of symbols 

# (these transformations must come in the script before the other

# transformation because the other are more general

# and would match these special cases as well.)

# Sometimes I wrote the phonetic variants of sʹ, zʹ:

# (FIXME: when I should use sʹ and ś in the recoded script?)

# (s^' can also represent Russian s^', as in 'es^'o'.)

s|s'^|ś|g

s|s^'|ś|g

s|z'^|ź|g

s|z^'|ź|g

# possibly I could write the phonetic variant of t':

s|c^'|ć|g # soft, of course

s|c'^|ć|g

# FIXME: How did I write an analoguous variant of d'? E.g., the word for \`Теньгушево' after a voiced? Or it is hard?

################################################

# Fricatives:

s|s^|š|g # fricative

s|c^|č|g # affricate, pronounced hard

s|z^|ž|g # fricative

s|Z^|ǯ|g # affricate

# At last, the simplest combinations:

#s|Z|ʒ|g # affricate -- Commented out, because there seems there weren't this sound in Mordvin or Komi speech, but I wrote it by mistake. So this conversion will catch the errors.

# (FIXME: What about affricates with the palatal sibilants: ź, ś? Z^'; and c^' is already present.

###############################################

# Vowels:

s|@|ə|g # schwa (unclear vowel that can be pronounced by them with any quality)

# FIXME: (Komi) try to put the Cyrillic-o-umlaut in the Cyrillic orthography; example:

#  казьт воспоминание казьт@д

#

###############################################

# The voiceless:

s|R|rˣ|g # the voiceless r

s|L|lˣ|g # the voiceless l

#s|F|wˣ|g # the voiceless w (sounds also like x, h)

# -- disabled 'F' because of the '\Fut' bug. Actually,

# 'W' for this sound looks more systematic in my system, so perhaps

# I should have used it, shouldn't I?

s|W|wˣ(?)|g # FIXME: sometimes I wrote 'W', is it the same sound?

s|J|jˣ|g # the voiceless j (sounds also like x, h; soft? Perhaps, I didn't distinguish 2 sounds--soft and hard--when using this sign.)

##############################################

# N:

s|N|ŋ|g # velar or nasalization

h # Save before the destructive checks (copy pattern space to hold space).

##########################################

# Well-formedness of the input (is checked destructively):

# - no ^'s should remain;

# - other capitals should not have been used (no S^, no C^ and no plain capitals);

# - more?

s|\$Date.*\$|| # ignore the 'preamble'

s|\\url{.*}|| # ignore filenames, which have been marked up like this: \url{Komi_rus_slovar1.pdf}

/\^/ { 

w /dev/stderr

q 56;} # (mostly GNU sed extensions)

/\(^\|[^\\]\)[[:upper:]]/ { # [^\\] -- a workaround for grammatical terms,

# to escape these restrictions on Latin characters: write them like \Pres, \Gen etc.

w /dev/stderr

q 57;} # (mostly GNU sed extensions)

g # Restore after the destructive checks (copy/append hold space to pattern space).

"

# FIXME: A "keep-going" mode: detect all the non-well-formednesses, and accumulate all

# the error codes in the exitCode (say, as bits).

exitCode="$?"

case "$exitCode" in

56) echo $"-- error: non well-formed input (a left-over '^')!" >/dev/stderr

    ;;

57) echo $"-- error: non well-formed input (a left-over capital Latin char)!" >/dev/stderr

    ;;

esac

exit "$exitCode"

# (Mordvin) Check that I have marked softenss? E.g., :

# |ti|t'i|