Convert long number into abbreviated string in JavaScript, with a special shortness requirement

I believe ninjagecko's solution doesn't quite conform with the standard you wanted. The following function does:

function intToString (value) {
    var suffixes = ["", "k", "m", "b","t"];
    var suffixNum = Math.floor((""+value).length/3);
    var shortValue = parseFloat((suffixNum != 0 ? (value / Math.pow(1000,suffixNum)) : value).toPrecision(2));
    if (shortValue % 1 != 0) {
        shortValue = shortValue.toFixed(1);
    }
    return shortValue+suffixes[suffixNum];
}

For values greater than 99 trillion no letter will be added, which can be easily fixed by appending to the 'suffixes' array.

Edit by Philipp follows: With the following changes it fits with all requirements perfectly!

function abbreviateNumber(value) {
    var newValue = value;
    if (value >= 1000) {
        var suffixes = ["", "k", "m", "b","t"];
        var suffixNum = Math.floor( (""+value).length/3 );
        var shortValue = '';
        for (var precision = 2; precision >= 1; precision--) {
            shortValue = parseFloat( (suffixNum != 0 ? (value / Math.pow(1000,suffixNum) ) : value).toPrecision(precision));
            var dotLessShortValue = (shortValue + '').replace(/[^a-zA-Z 0-9]+/g,'');
            if (dotLessShortValue.length <= 2) { break; }
        }
        if (shortValue % 1 != 0)  shortValue = shortValue.toFixed(1);
        newValue = shortValue+suffixes[suffixNum];
    }
    return newValue;
}

This handles very large values as well and is a bit more succinct and efficient.

abbreviate_number = function(num, fixed) {
  if (num === null) { return null; } // terminate early
  if (num === 0) { return '0'; } // terminate early
  fixed = (!fixed || fixed < 0) ? 0 : fixed; // number of decimal places to show
  var b = (num).toPrecision(2).split("e"), // get power
      k = b.length === 1 ? 0 : Math.floor(Math.min(b[1].slice(1), 14) / 3), // floor at decimals, ceiling at trillions
      c = k < 1 ? num.toFixed(0 + fixed) : (num / Math.pow(10, k * 3) ).toFixed(1 + fixed), // divide by power
      d = c < 0 ? c : Math.abs(c), // enforce -0 is 0
      e = d + ['', 'K', 'M', 'B', 'T'][k]; // append power
  return e;
}

Results:

for(var a='', i=0; i < 14; i++){ 
    a += i; 
    console.log(a, abbreviate_number(parseInt(a),0)); 
    console.log(-a, abbreviate_number(parseInt(-a),0)); 
}

0 0
-0 0
01 1
-1 -1
012 12
-12 -12
0123 123
-123 -123
01234 1.2K
-1234 -1.2K
012345 12.3K
-12345 -12.3K
0123456 123.5K
-123456 -123.5K
01234567 1.2M
-1234567 -1.2M
012345678 12.3M
-12345678 -12.3M
0123456789 123.5M
-123456789 -123.5M
012345678910 12.3B
-12345678910 -12.3B
01234567891011 1.2T
-1234567891011 -1.2T
0123456789101112 123.5T
-123456789101112 -123.5T
012345678910111213 12345.7T
-12345678910111212 -12345.7T

A lot of answers on this thread get rather complicated, using Math objects, map objects, for-loops, etc. But those approaches don't actually improve the design very much - they introduce more lines of code, more complexity, and more memory overhead. After evaluating several approaches, I think the manual approach is the easiest to understand, and provides the highest performance.

const formatCash = n => {
  if (n < 1e3) return n;
  if (n >= 1e3 && n < 1e6) return +(n / 1e3).toFixed(1) + "K";
  if (n >= 1e6 && n < 1e9) return +(n / 1e6).toFixed(1) + "M";
  if (n >= 1e9 && n < 1e12) return +(n / 1e9).toFixed(1) + "B";
  if (n >= 1e12) return +(n / 1e12).toFixed(1) + "T";
};

console.log(formatCash(1235000));