Angular Input Restriction Directive - Negating Regular Expressions

Question

EDIT: Please feel free to add additional validations that would be useful for others, using this simple directive.

--

I'm trying to create an Angular Directive that limits the characters input into a text box. I've been successful with a couple common use cases (alphbetical, alphanumeric and numeric) but using popular methods for validating email addresses, dates and currency I can't get the directive to work since I need it negate the regex. At least that's what I think it needs to do.

Any assistance for currency (optional thousand separator and cents), date (mm/dd/yyyy) and email is greatly appreciated. I'm not strong with regular expressions at all.

Here's what I have currently: http://jsfiddle.net/corydorning/bs05ys69/

HTML

<div ng-app="example">
<h1>Validate Directive</h1>

<p>The Validate directive allow us to restrict the characters an input can accept.</p>

<h3><code>alphabetical</code> <span style="color: green">(works)</span></h3>
<p>Restricts input to alphabetical (A-Z, a-z) characters only.</p>
<label><input type="text" validate="alphabetical" ng-model="validate.alphabetical"/></label>

<h3><code>alphanumeric</code> <span style="color: green">(works)</span></h3>
<p>Restricts input to alphanumeric (A-Z, a-z, 0-9) characters only.</p>
<label><input type="text" validate="alphanumeric" ng-model="validate.alphanumeric" /></label>

<h3><code>currency</code> <span style="color: red">(doesn't work)</span></h3>
<p>Restricts input to US currency characters with comma for thousand separator (optional) and cents (optional).</p>
<label><input type="text" validate="currency.us" ng-model="validate.currency" /></label>

<h3><code>date</code> <span style="color: red">(doesn't work)</span></h3>
<p>Restricts input to the mm/dd/yyyy date format only.</p>
<label><input type="text" validate="date" ng-model="validate.date" /></label>

<h3><code>email</code> <span style="color: red">(doesn't work)</span></h3>
<p>Restricts input to email format only.</p>
<label><input type="text" validate="email" ng-model="validate.email" /></label>

<h3><code>numeric</code> <span style="color: green">(works)</span></h3>
<p>Restricts input to numeric (0-9) characters only.</p>
<label><input type="text" validate="numeric" ng-model="validate.numeric" /></label>

JavaScript

angular.module('example', [])
  .directive('validate', function () {
    var validations = {
      // works
      alphabetical: /[^a-zA-Z]*$/,

      // works
      alphanumeric: /[^a-zA-Z0-9]*$/,

      // doesn't work - need to negate?
      // taken from: http://stackoverflow.com/questions/354044/what-is-the-best-u-s-currency-regex
      currency: /^[+-]?[0-9]{1,3}(?:,?[0-9]{3})*(?:\.[0-9]{2})?$/,

      // doesn't work - need to negate?
      // taken from here: http://stackoverflow.com/questions/15196451/regular-expression-to-validate-datetime-format-mm-dd-yyyy
      date: /(?:0[1-9]|1[0-2])\/(?:0[1-9]|[12][0-9]|3[01])\/(?:19|20)[0-9]{2}/,

      // doesn't work - need to negate?
      // taken from: http://stackoverflow.com/questions/46155/validate-email-address-in-javascript
      email: /^([\w-]+(?:\.[\w-]+)*)@((?:[\w-]+\.)*\w[\w-]{0,66})\.([a-z]{2,6}(?:\.[a-z]{2})?)$/i,

      // works
      numeric: /[^0-9]*$/
    };

  return {
    require: 'ngModel',

    scope: {
      validate: '@'
    },

    link: function (scope, element, attrs, modelCtrl) {
      var pattern = validations[scope.validate] || scope.validate
      ;

      modelCtrl.$parsers.push(function (inputValue) {
        var transformedInput = inputValue.replace(pattern, '')
        ;

        if (transformedInput != inputValue) {
          modelCtrl.$setViewValue(transformedInput);
          modelCtrl.$render();
        }

        return transformedInput;
      });
    }
  };
});

Answer 1

I am pretty sure, there is better way, probably regex is also not best tool for that, but here is mine proposition.

This way you can only restrict which characters are allowed for input and to force user to use proper format, but you will need to also validate final input after user will finish typing, but this is another story.

The alphabetic, numeric and alphanumeric are quite simple, for input and validating input, as it is clear what you can type, and what is a proper final input. But with dates, mails, currency, you cannot validate input with regex for full valid input, as user need to type it in first, and in a meanwhile the input need to by invalid in terms of final valid input. So, this is one thing to for example restrict user to type just digits and / for a date format, like: 12/12/1988, but in the end you need to check if he typed proper date or just 12/12/126 for example. This need to be checked when answer is submited by user, or when text field lost focus, etc.

To just validate typed character, you can try with this:

JSFiddle DEMO

First change:

var transformedInput = inputValue.replace(pattern, '')

to

var transformedInput = inputValue.replace(pattern, '$1')

then use regular expressions:

• /^([a-zA-Z]*(?=[^a-zA-Z]))./ - alphabetic
• /^([a-zA-Z0-9]*(?=[^a-zA-Z0-9]))./ - alphanumeric
• /(\.((?=[^\d])|\d{2}(?![^,\d.]))|,((?=[^\d])|\d{3}(?=[^,.$])|(?=\d{1,2}[^\d]))|\$(?=.)|\d{4,}(?=,)).|[^\d,.$]|^\$/- currency (allow string like: 343243.34, 1,123,345.34, .05 with or without $)
• ^(((0[1-9]|1[012])|(\d{2}\/\d{2}))(?=[^\/])|((\d)|(\d{2}\/\d{2}\/\d{1,3})|(.+\/))(?=[^\d])|\d{2}\/\d{2}\/\d{4}(?=.)).|^(1[3-9]|[2-9]\d)|((?!^)(3[2-9]|[4-9]\d)\/)|[3-9]\d{3}|2[1-9]\d{2}|(?!^)\/\d\/|^\/|[^\d/] - date (00-12/00-31/0000-2099)
• /^(\d*(?=[^\d]))./ - numeric
• /^([\w.$-]+\@[\w.]+(?=[^\w.])|[\w.$-]+\@(?=[^\w.-])|[\w.@-]+(?=[^\w.$@-])).$|\.(?=[^\w-@]).|[^\w.$@-]|^[^\w]|\.(?=@).|@(?=\.)./i - email

Generally, it use this pattern:

([valid characters or structure] captured in group $1)(?= positive lookahead for not allowed characters) any character

in effect it will capture all valid character in group $1, and if user type in an invalid character, whole string is replaced with already captured valid characters from group $1. It is complemented by part which shall exclude some obvious invalid character(s), like @@ in a mail, or 34...2 in currency.

With understanding how these regular expression works, despite that it looks quite complex, I think it easy to extend it, by adding additional allowed/not allowed characters.

Regular expression for validating currency, dates and mails are easy to find, so I find it redundant to post them here.

OffTopic. Whats more the currency part in your demo is not working, it is bacause of: validate="currency.us" instead of validate="currency", or at least it works after this modification.

Answer 2

In my opinion it is impossible to create regular expressions that will work for matching things like dates or emails with the parser you use. This is mainly because you would need non-capturing groups in your regular expressions (which is possible), which are not replaced by the inputValue.replace(pattern, '') call you have in your parser function. And this is the part that is not possible in JavaScript. JavaScript replaces what you put in non-capturing groups as well.

So... you'll need to go for a different approach. I would suggest to go for positive regular expressions, which will yield a match when the input is valid. Then you need of course to change the code of your parser. You could for instance decide to chop off characters from the end of the input text until what remains passes the regular expression test. This you could code as follows:

  modelCtrl.$parsers.push(function (inputValue) {
    var transformedInput = inputValue;
    while (transformedInput && !pattern.exec(transformedInput)) {
       // validation fails: chop off last character and try again
       transformedInput = transformedInput.slice(0, -1);
    }

    if (transformedInput !== inputValue) {
      modelCtrl.$setViewValue(transformedInput);
      modelCtrl.$render();
    }

    return transformedInput;
  });

Now life has become a bit easier. Just pay attention that you make your regular expressions in such a way that they do not reject partial input. So "01/" should be considered valid for a date, otherwise the user can never get to type in a date. On the other hand, as soon as it becomes clear that adding characters will no longer allow for valid input, the regular expression should reject it. So "101" should be rejected as a date, as you can never add characters at the end to make it a valid date.

Also, all of these regular expressions should check the whole input, so as a consequence they need to make use of the ^ and $ symbols.

Here is what the regular expression for a (partial) date could look like:

^([0-9]{0,2}|[0-9]{2}[\/]([0-9]{0,2}|[0-9]{2}[\/][0-9]{0,4}))$

This means: an input of 0 to 2 digits is valid, or exactly 2 digits followed by a slash, followed by either:

1. 0 to 2 digits, or
2. exactly 2 digits followed by a slash, followed by 0 to 4 digits

Admittedly, not as smart as the one you had found, but that one would need a lot of editing to allow for partially entered dates. It is possible, but it represents a very long expression with a lot of brackets and |.

Once you have all the regular expressions set up, you could think to further improve the parser. One idea would be to not let it chop off characters from the end, but to let it test all strings with one character removed somewhere compared to the original, and see which one passes the test. If there is no way found to remove one character and have success, then remove two consecutive characters in any place of the input value, then three, ... etc, until you find a value that passes the test or arrive at an empty value.

This will work better for cases where the user inserts characters half way their input. Just an idea...