5310 / ubn2rbn Goto Github PK

That "Please wait" thingy doesn't go away even after romanisation is complete.

https://imgur.com/hQ8zz&caBSj

'য্য-ফলা' is written as '্য', which is included into the char-map on virtue or appearing to be a single character. But in reality, it is two characters, not a ligature by default.

This is why when converting words such as 'মধ্যে', the 'য' is converted to 'j' first, as it appears before '্য' in the map.

The proper way to fix this would be to include '্য' in the comp-map, along with all such mappings which are more than one letter.

All Vowel-mix-ins also fall into this category, but they are ligatures, which is why they work. Perhaps we should let them be.

But that's not all, folks: #7 (comment)

To be honest, I can't get it to work.

The desired behavior is to try and convert the selected text; if there is no selected text, then convert the entire page (or text-area, if it is one.)

But I just can't get wrapSelection to work, even during test. I must be doing something obvious wrong, so I'll look into it. But if I still can't get it to work, I will probably do it the old-fashioned way...

Subsetting works, but instead of only styling the glyphs falling under the Bengali group, it styles any line with Bengali text, including any English ones, while leaving the others clear.

Tested only in Chrome, and only from the locally installed font.

Make a minified version of the bookmarklet, and link to that.

The readme should actually be useful.

And do make a habit of adding an explicit license.

Currently, there's a very noticeable pause between the invocation of the bookmarklet (link) and actual change.

People, myself included, often think that the bookmarklet is broken, or not loading.

A simple notification, a throbber, or a notice, will be very helpful.

The demo should have some minimal instruction like "drag the above text to our bookmark-bar."

The demo should also have a tag for Github, and a link-back to the repo.

As noted in #16, this regex is changing something it should, and the resulting markup is broken.

While rife with exceptions, that routine is important. Please fix.

Sentence-case does not always work. This is due to the fact that raw-HTML is messy and I've had enough regex.

@Oni64 has already solved the issue of whether to append a ্ to a word-endning unmodified consonant with absolute minimum of exceptions.

But I can't seem to implement it inside the 'let.

This jsFiddle works perfectly well. It only adds the ending ্, but does not romanize, of course. But the same thing does not work inside the bookmarklet.

Furthermore, since complex regex as string is quite annoying to make with the RegExp() constructor, I'm not using the usual method of creating a dictionary, instead, appending the text manually. Which is acceptable, but was worth mentioning.

Algorith does not yet allow for default vowel-sound for consonants, all unmodified consonants map to the non-vowel sound.

There's no 'kô', only 'k'.

I need help completing the dictionaries containing romanization-maps.

To begin with, only charmap would do.

The first letter of every new line should be capitalized, like English.

All characters that are a composite of a plain character and the dot-glyph (়) are parsed individually instead of by the ligature.

So far, I've only noticed to 'য়' behave like it's 'য+়'

Most unmodified consonants are indeed pronounced with 'ô', but some are unvoiced.

@Oni64 has actually worked very thoroughly on this issue. While he has come up with an model that works for a lot of cases, it has a much lower adherence/exception ratio than the word-ending one.

The only reliable way to solve that problem would be to include a respectable phrase-map. Procedurally, this has already been implemented, but the phrase-map does have such width.

We have decided to not try to solve this by algorithm after all. And will look into whether a useful phrase-map can be built without too much bloat. Or we could leave things as is, since it's already quite readable.