Detecting if a character is a letter

Question

Given a set of words, I need to put them in an hash keyed on the first letter of the word. I have words = {}, with keys A..Z and 0 for numbers and symbols. I was doing something like

var firstLetter = name.charAt(0);
    firstLetter = firstLetter.toUpperCase();

if (firstLetter < "A" || firstLetter > "Z") {
    firstLetter = "0";
}
if (words[firstLetter] === undefined) {
    words[firstLetter] = [];
} 
words[firstLetter].push(name);

but this fails with dieresis and other chars, like in the word Ärzteversorgung. That word is put in the "0" array, how could I put it in the "A" array?

Do you only want to have characters like Ä detected as letters, or do you want to have Ä detected as if it were an A? — Bergi, May 22 '12 at 18:37
Ä is not an A. You will need a mapping of characters with accents to without accents. — Starkey, May 22 '12 at 18:37
you have to map for Ä and other alphabets of this character too. like you are doing for 0 1 2 3..... and abc etc. — Rizstien, May 22 '12 at 18:41
Check out this post http://stackoverflow.com/questions/863800/replacing-diacritics-in-javascript — Heitor Chang, May 22 '12 at 18:42
Would using a regex test, like `/[\w\u00A0-\uD7FF\uF900-\uFDCF\uFDF0-\uFFEF]/.test(firstLetter)` be appropriate? — Chris Carew, May 22 '12 at 18:46
@HeitorChang this duplicate have better answers http://stackoverflow.com/questions/990904/javascript-remove-accents-in-strings — ajax333221, May 22 '12 at 19:02
I need to map Ä as A. The link suggested by ajax333221 has answers that solve to my question, as you have to transform the string using some kind of map — cdarwin, May 22 '12 at 20:13
you may want to remove accents & then do a simple [a-z] check. see http://stackoverflow.com/questions/990904/javascript-remove-accents-in-strings — Adriano, Sep 15 '14 at 13:35

score 16 · Answer 1 · answered May 22 '12 at 18:42

16

You can use this to test if a character is likely to be a letter:

var firstLetter = name.charAt(0).toUpperCase();
if( firstLetter.toLowerCase() != firstLetter) {
    // it's a letter
}
else {
    // it's a symbol
}

This works because JavaScript already has a mapping for lowercase to uppercase letters (and vice versa), so if a character is unchanged by toLowerCase() then it's not in the letter table.

answered May 22 '12 at 18:42

Niet the Dark Absol

320,036
81
464
592

1

An interesting trick, but as you emphasize, it works just “likely”. But it might be used if you add ad hoc checks for all the characters that may appear and that cause a wrong result in the simple test. In the Latin 1 range, the following characters get misclassified: º and ª (masculine and feminine ordinal indicators), sharp ß (probably the most relevant character here), and debatably the micro sign µ (formally a letter in Unicode, compatibility equivalent to Greek letter mu, but widely understood as a special character rather than a letter). – Jukka K. Korpela May 22 '12 at 19:07
4

It only works for characters in bicameral scripts, i.e. writing systems that make uppercase/lowercase distinction; most scripts don’t (e.g., Hebrew, Devanagari, Chinese). – Jukka K. Korpela Nov 03 '12 at 14:02
@JukkaK.Korpela - Yes there are cons and the pros are the speed of the check. this can processed much faster than anything else, and English will likely (should) be the only language most people will need) – vsync Jan 13 '14 at 00:44
3

@vsync, the question mentions the sample word “Ärzteversorgung”. It isn’t English. Typically if people only think of English, they don’t even ask this question—they just assume that `[A-Za-z]` covers all letters. – Jukka K. Korpela Jan 13 '14 at 09:46
Questions in Stackoverflow sometimes means nothing because I came here from Google and what the OP asked for is not what the title suggests, therefore it's important to cover the answers for the people who do come here from Google for answers. – vsync Jan 13 '14 at 13:40
As of 2020: In a latin script you can use the uppercase/lowercase comparison to mark a character as a letter. Try this: 'ß'.toUpperCase() // ==> 'SS', 'ſ'.toUpperCase() // ==> 'S'. Even in Greek: 'µ'.toUpperCase() // ==> 'Μ' (for \u00B5 as well as \u03BC). Only for the ordinal indicators the lowercase === uppercase. – Onno van der Zee Apr 13 '20 at 22:29

score 5 · Answer 2 · answered Aug 19 '14 at 11:22

5

Try converting the character to its uppercase and lowercase and check to see if there's a difference. Only letter characters change when they are converted to their respective upper and lower case (numbers, punctuation marks, etc. don't). Below is a sample function using this concept in mind:

function isALetter(charVal)
{
    if( charVal.toUpperCase() != charVal.toLowerCase() )
       return true;
    else
       return false;
}

answered Aug 19 '14 at 11:22

JDE

59
1
2

This does not work. The expression `"Á" == "á"` will always return `false`! – Javi Marzán Aug 21 '20 at 12:38

score 4 · Accepted Answer · answered May 22 '12 at 18:45

You could use a regular expression. Unfortunately, JavaScript does not consider international characters to be "word characters". But you can do it with the regular expression below:

var firstLetter = name.charAt(0);
firstLetter = firstLetter.toUpperCase();
if (!firstLetter.match(/^\wÀÈÌÒÙàèìòùÁÉÍÓÚÝáéíóúýÂÊÎÔÛâêîôûÃÑÕãñõÄËÏÖÜäëïöüçÇßØøÅåÆæÞþÐð$/)) {
    firstLetter = "0";
}
if (words[firstLetter] === undefined) {
    words[firstLetter] = [];
} 
words[firstLetter].push(name);

score 2 · Answer 4 · answered May 22 '12 at 18:44

2

You can use .charCodeAt(0); to get the position in the ASCII Chart and then do some checks.

The ranges you are looking for are probably 65-90, 97-122, 128-154, 160-165 (inclusive), but double check this by viewing the ASCII Chart

Something like this

if((x>64&&x<91)||(x>96&&x<123)||(x>127&&x<155)||(x>159&&x<166))

Where x is the Char Code

answered May 22 '12 at 18:44

ajax333221

11,436
16
61
95

I don't know, [charCodeAt MDN](https://developer.mozilla.org/en/JavaScript/Reference/Global_Objects/String/charCodeAt) says _"Unicode code points range from 0 to 1,114,111"_ – ajax333221 May 22 '12 at 18:57
@jnrbsn No it won't... ASCII does not cover for all of what is considered a letter in every language – Jérôme Verstrynge Jun 13 '13 at 21:33
@ajax333221 There are some valid code points which are not printable characters – Jérôme Verstrynge Jun 13 '13 at 21:34
this works for bytes, but not for characters... a character can be more than one byte – milahu Aug 31 '23 at 11:24

score 2 · Answer 5 · answered Mar 24 '20 at 08:17

This is fortunately now possible without external libraries. Straight from the docs:

let story = "It’s the Cheshire Cat: now I shall have somebody to talk to.";

// Most explicit form
story.match(/\p{General_Category=Letter}/gu);

// It is not mandatory to use the property name for General categories
story.match(/\p{Letter}/gu);

Detecting if a character is a letter

5 Answers5

Linked