transliterator as a morphology processor?

Recently been doing some work with charset_table to allow users to search for 'accented' letters more easily, folding them in similar way to uppercase, so 'accent insensitive'. ie U+C3->a sort of thing.

But a question was raised, what about when want to search for a specific variation, eg specifically search for "Bryn Côch" vs "Bryn Coch".

using index_exact_words springs to mind, but as will have already been processed by charset_table it doesn't work.

Wonder if any mileage, in being able to do the transliteration (to ASCII) via morphology. So that exact form operator would get the raw version.

iconv etc has well known transliteration tables, that could be used. https://gist.github.com/paulsmith/6059259

Only wrinkle morphology preprocessors tend to 'abort' when changing the token, si it would have to be hardwired, so that following processors (like stemming) could still be used.

morphology = translit_ascii, stem_en

(charset_table would have to be 'open' to let them though as word-chars, but wouldn't do the transliteration itself)

Start by tracing the morphology and charset_table processing described in the issue, including how index_exact_words preserves exact forms. Assess the proposed translit_ascii processor and its interaction with stem_en and word-character handling. Done means a decided design and implementation path for ASCII transliteration without preventing exact-form searches.