@hscii/htrlib
v1.1.1
Published
e52 is English (26+26)
Readme
e52 is English (26+26)
u9 : 9 indian writing scripts u10 : 9 indian writing scripts + 1 srilanka writing script
e52 is English (26+26) e23 is English(abcdefghi klmnop rstu wxyz) xe38 : xnglo_english 38 alphabets ( 26 a-z + 11 soft konsonants K G C Z T D J Q B S + velar nasal N always has 1 k/K/g/G next like kiNg , riNg etc) u9 : 9 indian writing scripts u10 : 9 indian writing scripts + 1 srilanka writing script
list of xnglo languages : xe38(xnglo_english_in_38_chars) xv38(xnglo_hindi) , xb38(xnglo_bengali), xmr38(xnglo_mrathi) , xg38(xnglo_guzrati), xp38(xnglo_pnzabi) , xo38(xnglo_odia) , xj38(xnglo_telugu) , xt38(xnglo_tamil), xm38(xnglo_malayalam) , xk38(xnglo_kannada) , xs38(xnglo_sinhala), xi38(xnglo_india=xv38/xb38/xmr38/xp38/xo38/xj38/xt38/xm38/xk38/xg38/xs38 is xnglo script for all 10 indian/srilankan languages.)
csv for xnglo phonetics (phonemes/graphemes) is ./xNglo_phoneme_grapheme.csv xnglo script for all xnglo langauges is made up from albhabets present in ./xNglo_phoneme_grapheme.csv these are 38 : a-z K G C Z T Q J Q N P B R S
this project is to create npmjs typescript library for transformation phrom/tu : phrom(e52/u10/e52_u10) -> tu(e23/xe/xi/xv/xb/xp/xo/xj/xt/xm/xk/xs/xmr)
for e52->e23 there should be function e52_tu_e23. it should do transliterate as replace("lover", "lwxr"). replace("never", "nxwxr"). replace("vest", "weist"). replace("vine", "wayin"). replace("vary", "wxyri"). replace("vet", "wyt"). replace("v", "w"). replace("j", "z"). replace("q", "k")
for u9 to xi38: u9_to_xi38 character wise transliteration function based on mappings in ./u9_to_xi38.csv
for unicode_sinhala to xi38: unicode_sinhala_to_xi38 character wise transliteration function based on mappings in ./u9_to_xi38.csv
for u10 to xi38: u10_to_xi38 will call u9_to_xi38/unicode_sinhala_to_xi38 character wise transliteration function
for e52->xe38 :
- e52->transliterate->punjabi('pa')->u10_to_xi->xe38
for e52 to xv38/xb38/xp38/xo38/xj38/xt38/xm38/xk38/xs38/xmr38
e52 to hindi/bengali/punzabi/odia/telugu/tamil/kannada/sinhala/marathi google translation.
translated text -> u10_to_xi -> xv/xb/xp/xo/xj/xt/xm/xk/xs/xmr
du changes and test
- write tests to test chages in sometest.test.ts file and sAve in tests tree.
- npm install (after running qhis "htrlib\node_modules" dir gets created so gitignore node_modules
- npm run test (now tests dir not present so skip qhis)
if tests passed then :
- npm login
- npm version patch // --no-git-tag-version && npm publish --access public (# Instead of patch, you can use 'minor' or 'major')
- npm publish
namespace hsciistr {
export enum enum_from { ascii = 'ascii', indikunicode = 'indikunicode', ascii_and_indik = 'ascii_and_indik' };
export enum enum_tu {
all='all', inglish='inglish', korean='korean', russian='russian', hindi='hindi', bangla='bangla', gurmukhi='gurmukhi',
guzrati='guzrati', oriya='oriya', tamil='tamil', kannada='kannada', telugu='telugu', malayalam='malayalam', sinhala='sinhala'
};
}
class hsciistr {
istr: string;
ostrdict: { [key: string]: string } = {
inglish: '', korean: '', russian: '', hindi: '', bangla: '', gurmukhi: '', guzrati: '', oriya: '', tamil: '', kannada: '', telugu: '', malayalam: '', sinhala: ''
};
constructor(from=hsciistr.enum_from.ascii_and_indik, tu=hsciistr.enum_tu.all) {
this.from = from ;
this.tu = tu ;
this.istr = "";
}
from: hsciistr.enum_from;
tu: hsciistr.enum_tu;
setistr(istr: string): hsciistr { this.istr = istr; return this; }
setfromstr(fromstr: hsciistr.enum_from): hsciistr { this.from = fromstr; return this; }
settostr(tostr: hsciistr.enum_tu): hsciistr { this.tu = tostr; return this; }
duztr(): hsciistr { /** transliteration code */ return this; }
npm login details : hscii / punxmrazxn= . i will get otp also. https://addons.mozilla.org/en-US/firefox/users/edit : v..g_atrate_gmail.com name is hscii https://addons.mozilla.org/en-US/firefox/user/17735318/ extensions https://addons.mozilla.org/en-US/firefox/addon/linguist84/
hsciistr => open hsciistr.md
hscii npmjs packages url : https://www.npmjs.com/settings/hscii/packages htrlib libhsciifp
