mirror of https://github.com/zulip/zulip.git
297 lines
8.7 KiB
JavaScript
297 lines
8.7 KiB
JavaScript
var util = (function () {
|
|
|
|
var exports = {};
|
|
|
|
// From MDN: https://developer.mozilla.org/en-US/docs/JavaScript/Reference/Global_Objects/Math/random
|
|
exports.random_int = function random_int(min, max) {
|
|
return Math.floor(Math.random() * (max - min + 1)) + min;
|
|
};
|
|
|
|
// Like C++'s std::lower_bound. Returns the first index at which
|
|
// `value` could be inserted without changing the ordering. Assumes
|
|
// the array is sorted.
|
|
//
|
|
// `first` and `last` are indices and `less` is an optionally-specified
|
|
// function that returns true if
|
|
// array[i] < value
|
|
// for some i and false otherwise.
|
|
//
|
|
// Usage: lower_bound(array, value, [less])
|
|
// lower_bound(array, first, last, value, [less])
|
|
exports.lower_bound = function (array, arg1, arg2, arg3, arg4) {
|
|
var first;
|
|
var last;
|
|
var value;
|
|
var less;
|
|
if (arg3 === undefined) {
|
|
first = 0;
|
|
last = array.length;
|
|
value = arg1;
|
|
less = arg2;
|
|
} else {
|
|
first = arg1;
|
|
last = arg2;
|
|
value = arg3;
|
|
less = arg4;
|
|
}
|
|
|
|
if (less === undefined) {
|
|
less = function (a, b) { return a < b; };
|
|
}
|
|
|
|
var len = last - first;
|
|
var middle;
|
|
var step;
|
|
while (len > 0) {
|
|
step = Math.floor(len / 2);
|
|
middle = first + step;
|
|
if (less(array[middle], value, middle)) {
|
|
first = middle;
|
|
first += 1;
|
|
len = len - step - 1;
|
|
} else {
|
|
len = step;
|
|
}
|
|
}
|
|
return first;
|
|
};
|
|
|
|
exports.same_stream_and_topic = function util_same_stream_and_topic(a, b) {
|
|
// Streams and topics are case-insensitive.
|
|
return a.stream_id === b.stream_id &&
|
|
a.subject.toLowerCase() === b.subject.toLowerCase();
|
|
};
|
|
|
|
exports.is_pm_recipient = function (email, message) {
|
|
var recipients = message.reply_to.toLowerCase().split(',');
|
|
return recipients.indexOf(email.toLowerCase()) !== -1;
|
|
};
|
|
|
|
exports.extract_pm_recipients = function (recipients) {
|
|
return _.filter(recipients.split(/\s*[,;]\s*/), function (recipient) {
|
|
return recipient.trim() !== "";
|
|
});
|
|
};
|
|
|
|
exports.same_recipient = function util_same_recipient(a, b) {
|
|
if (a === undefined || b === undefined) {
|
|
return false;
|
|
}
|
|
if (a.type !== b.type) {
|
|
return false;
|
|
}
|
|
|
|
switch (a.type) {
|
|
case 'private':
|
|
if (a.to_user_ids === undefined) {
|
|
return false;
|
|
}
|
|
return a.to_user_ids === b.to_user_ids;
|
|
case 'stream':
|
|
return exports.same_stream_and_topic(a, b);
|
|
}
|
|
|
|
// should never get here
|
|
return false;
|
|
};
|
|
|
|
exports.same_sender = function util_same_sender(a, b) {
|
|
return a !== undefined && b !== undefined &&
|
|
a.sender_email.toLowerCase() === b.sender_email.toLowerCase();
|
|
};
|
|
|
|
exports.normalize_recipients = function (recipients) {
|
|
// Converts a string listing emails of message recipients
|
|
// into a canonical formatting: emails sorted ASCIIbetically
|
|
// with exactly one comma and no spaces between each.
|
|
recipients = _.map(recipients.split(','), function (s) { return s.trim(); });
|
|
recipients = _.map(recipients, function (s) { return s.toLowerCase(); });
|
|
recipients = _.filter(recipients, function (s) { return s.length > 0; });
|
|
recipients.sort();
|
|
return recipients.join(',');
|
|
};
|
|
|
|
// Avoid URI decode errors by removing characters from the end
|
|
// one by one until the decode succeeds. This makes sense if
|
|
// we are decoding input that the user is in the middle of
|
|
// typing.
|
|
exports.robust_uri_decode = function (str) {
|
|
var end = str.length;
|
|
while (end > 0) {
|
|
try {
|
|
return decodeURIComponent(str.substring(0, end));
|
|
} catch (e) {
|
|
if (!(e instanceof URIError)) {
|
|
throw e;
|
|
}
|
|
end -= 1;
|
|
}
|
|
}
|
|
return '';
|
|
};
|
|
|
|
exports.rtrim = function (str) {
|
|
return str.replace(/\s+$/, '');
|
|
};
|
|
|
|
// If we can, use a locale-aware sorter. However, if the browser
|
|
// doesn't support the ECMAScript Internationalization API
|
|
// Specification, do a dumb string comparison because
|
|
// String.localeCompare is really slow.
|
|
exports.make_strcmp = function () {
|
|
try {
|
|
var collator = new Intl.Collator();
|
|
return collator.compare;
|
|
} catch (e) {
|
|
// continue regardless of error
|
|
}
|
|
|
|
return function util_strcmp(a, b) {
|
|
return a < b ? -1 : a > b ? 1 : 0;
|
|
};
|
|
};
|
|
exports.strcmp = exports.make_strcmp();
|
|
|
|
exports.escape_regexp = function (string) {
|
|
// code from https://developer.mozilla.org/en-US/docs/Web/JavaScript/Guide/Regular_Expressions
|
|
// Modified to escape the ^ to appease jslint. :/
|
|
return string.replace(/([.*+?\^=!:${}()|\[\]\/\\])/g, "\\$1");
|
|
};
|
|
|
|
exports.array_compare = function util_array_compare(a, b) {
|
|
if (a.length !== b.length) {
|
|
return false;
|
|
}
|
|
var i;
|
|
for (i = 0; i < a.length; i += 1) {
|
|
if (a[i] !== b[i]) {
|
|
return false;
|
|
}
|
|
}
|
|
return true;
|
|
};
|
|
|
|
/* Represents a value that is expensive to compute and should be
|
|
* computed on demand and then cached. The value can be forcefully
|
|
* recalculated on the next call to get() by calling reset().
|
|
*
|
|
* You must supply a option to the constructor called compute_value
|
|
* which should be a function that computes the uncached value.
|
|
*/
|
|
var unassigned_value_sentinel = {};
|
|
exports.CachedValue = function (opts) {
|
|
this._value = unassigned_value_sentinel;
|
|
_.extend(this, opts);
|
|
};
|
|
|
|
exports.CachedValue.prototype = {
|
|
get: function CachedValue_get() {
|
|
if (this._value === unassigned_value_sentinel) {
|
|
this._value = this.compute_value();
|
|
}
|
|
return this._value;
|
|
},
|
|
|
|
reset: function CachedValue_reset() {
|
|
this._value = unassigned_value_sentinel;
|
|
},
|
|
};
|
|
|
|
exports.is_all_or_everyone_mentioned = function (message_content) {
|
|
var all_everyone_re = /(^|\s)(@\*{2}(all|everyone|stream)\*{2})($|\s)/;
|
|
return all_everyone_re.test(message_content);
|
|
};
|
|
|
|
exports.move_array_elements_to_front = function util_move_array_elements_to_front(array, selected) {
|
|
var i;
|
|
var selected_hash = {};
|
|
for (i = 0; i < selected.length; i += 1) {
|
|
selected_hash[selected[i]] = true;
|
|
}
|
|
var selected_elements = [];
|
|
var unselected_elements = [];
|
|
for (i = 0; i < array.length; i += 1) {
|
|
if (selected_hash[array[i]]) {
|
|
selected_elements.push(array[i]);
|
|
} else {
|
|
unselected_elements.push(array[i]);
|
|
}
|
|
}
|
|
// Add the unselected elements after the selected ones
|
|
return selected_elements.concat(unselected_elements);
|
|
};
|
|
|
|
// check by the userAgent string if a user's client is likely mobile.
|
|
exports.is_mobile = function () {
|
|
var regex = "Android|webOS|iPhone|iPad|iPod|BlackBerry|IEMobile|Opera Mini";
|
|
return new RegExp(regex, "i").test(window.navigator.userAgent);
|
|
};
|
|
|
|
exports.prefix_sort = function (query, objs, get_item) {
|
|
// Based on Bootstrap typeahead's default sorter, but taking into
|
|
// account case sensitivity on "begins with"
|
|
var beginswithCaseSensitive = [];
|
|
var beginswithCaseInsensitive = [];
|
|
var noMatch = [];
|
|
var obj;
|
|
var item;
|
|
for (var i = 0; i < objs.length; i += 1) {
|
|
obj = objs[i];
|
|
if (get_item) {
|
|
item = get_item(obj);
|
|
} else {
|
|
item = obj;
|
|
}
|
|
if (item.indexOf(query) === 0) {
|
|
beginswithCaseSensitive.push(obj);
|
|
} else if (item.toLowerCase().indexOf(query.toLowerCase()) === 0) {
|
|
beginswithCaseInsensitive.push(obj);
|
|
} else {
|
|
noMatch.push(obj);
|
|
}
|
|
}
|
|
return {
|
|
matches: beginswithCaseSensitive.concat(beginswithCaseInsensitive),
|
|
rest: noMatch,
|
|
};
|
|
};
|
|
|
|
// manipulate prefix_sort to select popular emojis first
|
|
// This is kinda a hack and so probably not our long-term solution.
|
|
exports.emoji_prefix_sort = function (query, objs, get_item) {
|
|
var prefix_sort = exports.prefix_sort(query, objs, get_item);
|
|
var popular_emoji_matches = [];
|
|
var other_emoji_matches = [];
|
|
prefix_sort.matches.forEach(function (obj) {
|
|
if (emoji_picker.frequently_used_emojis_list.indexOf(obj.codepoint) !== -1) {
|
|
popular_emoji_matches.push(obj);
|
|
} else {
|
|
other_emoji_matches.push(obj);
|
|
}
|
|
});
|
|
return { matches: popular_emoji_matches.concat(other_emoji_matches), rest: prefix_sort.rest };
|
|
};
|
|
|
|
function to_int(s) {
|
|
return parseInt(s, 10);
|
|
}
|
|
|
|
exports.sorted_ids = function (ids) {
|
|
// This mapping makes sure we are using ints, and
|
|
// it also makes sure we don't mutate the list.
|
|
var id_list = _.map(ids, to_int);
|
|
id_list.sort(function (a, b) {
|
|
return a - b;
|
|
});
|
|
id_list = _.uniq(id_list, true);
|
|
|
|
return id_list;
|
|
};
|
|
|
|
return exports;
|
|
|
|
}());
|
|
if (typeof module !== 'undefined') {
|
|
module.exports = util;
|
|
}
|