aboutsummaryrefslogtreecommitdiffstats
path: root/lib/utils.coffee
blob: 45a880e5669381b679b0b9ea5e305d68fddc9cf0 (plain)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
# Only pass events to the handler if they are marked as trusted by the browser.
# This is kept in the global namespace for brevity and ease of use.
window.forTrusted ?= (handler) -> (event) ->
  if event?.isTrusted
    handler.apply this, arguments
  else
    true

browserInfo = browser?.runtime?.getBrowserInfo?()

Utils =
  isFirefox: do ->
    # NOTE(mrmr1993): This test only works in the background page, this is overwritten by isEnabledForUrl for
    # content scripts.
    isFirefox = false
    browserInfo?.then? (browserInfo) ->
      isFirefox = browserInfo?.name == "Firefox"
    -> isFirefox
  firefoxVersion: do ->
    # NOTE(mrmr1993): This only works in the background page.
    ffVersion = undefined
    browserInfo?.then? (browserInfo) ->
      ffVersion = browserInfo?.version
    -> ffVersion
  getCurrentVersion: ->
    chrome.runtime.getManifest().version

  # Returns true whenever the current page (or the page supplied as an argument) is from the extension's
  # origin (and thus can access the extension's localStorage).
  isExtensionPage: (win = window) -> try win.document.location?.origin + "/" == chrome.extension.getURL ""

  # Returns true whenever the current page is the extension's background page.
  isBackgroundPage: -> @isExtensionPage() and chrome.extension.getBackgroundPage?() == window

  # Escape all special characters, so RegExp will parse the string 'as is'.
  # Taken from http://stackoverflow.com/questions/3446170/escape-string-for-use-in-javascript-regex
  escapeRegexSpecialCharacters: do ->
    escapeRegex = /[\-\[\]\/\{\}\(\)\*\+\?\.\\\^\$\|]/g
    (str) -> str.replace escapeRegex, "\\$&"

  escapeHtml: (string) -> string.replace(/</g, "&lt;").replace(/>/g, "&gt;")

  # Generates a unique ID
  createUniqueId: do ->
    id = 0
    -> id += 1

  hasChromePrefix: do ->
    chromePrefixes = [ "about:", "view-source:", "extension:", "chrome-extension:", "data:" ]
    (url) ->
      for prefix in chromePrefixes
        return true if url.startsWith prefix
      false

  hasJavascriptPrefix: (url) ->
    url.startsWith "javascript:"

  hasFullUrlPrefix: do ->
    urlPrefix = new RegExp "^[a-z][-+.a-z0-9]{2,}://."
    (url) -> urlPrefix.test url

  # Decode valid escape sequences in a URI.  This is intended to mimic the best-effort decoding
  # Chrome itself seems to apply when a Javascript URI is enetered into the omnibox (or clicked).
  # See https://code.google.com/p/chromium/issues/detail?id=483000, #1611 and #1636.
  decodeURIByParts: (uri) ->
    uri.split(/(?=%)/).map((uriComponent) ->
      try
        decodeURIComponent uriComponent
      catch
        uriComponent
    ).join ""

  # Completes a partial URL (without scheme)
  createFullUrl: (partialUrl) ->
    if @hasFullUrlPrefix(partialUrl) then partialUrl else ("http://" + partialUrl)

  # Tries to detect if :str is a valid URL.
  isUrl: (str) ->
    # Must not contain spaces
    return false if ' ' in str

    # Starts with a scheme: URL
    return true if @hasFullUrlPrefix str

    # More or less RFC compliant URL host part parsing. This should be sufficient for our needs
    urlRegex = new RegExp(
      '^(?:([^:]+)(?::([^:]+))?@)?' + # user:password (optional) => \1, \2
      '([^:]+|\\[[^\\]]+\\])'       + # host name (IPv6 addresses in square brackets allowed) => \3
      '(?::(\\d+))?$'                 # port number (optional) => \4
      )

    # Official ASCII TLDs that are longer than 3 characters + inofficial .onion TLD used by TOR
    longTlds = ['arpa', 'asia', 'coop', 'info', 'jobs', 'local', 'mobi', 'museum', 'name', 'onion']

    specialHostNames = ['localhost']

    # Try to parse the URL into its meaningful parts. If matching fails we're pretty sure that we don't have
    # some kind of URL here.
    match = urlRegex.exec (str.split '/')[0]
    return false unless match
    hostName = match[3]

    # Allow known special host names
    return true if hostName in specialHostNames

    # Allow IPv6 addresses (need to be wrapped in brackets as required by RFC). It is sufficient to check for
    # a colon, as the regex wouldn't match colons in the host name unless it's an v6 address
    return true if ':' in hostName

    # At this point we have to make a decision. As a heuristic, we check if the input has dots in it. If yes,
    # and if the last part could be a TLD, treat it as an URL
    dottedParts = hostName.split '.'

    if dottedParts.length > 1
      lastPart = dottedParts.pop()
      return true if 2 <= lastPart.length <= 3 or lastPart in longTlds

    # Allow IPv4 addresses
    return true if /^(\d{1,3}\.){3}\d{1,3}$/.test hostName

    # Fallback: no URL
    return false

  # Map a search query to its URL encoded form. The query may be either a string or an array of strings.
  # E.g. "BBC Sport" -> "BBC+Sport".
  createSearchQuery: (query) ->
    query = query.split(/\s+/) if typeof(query) == "string"
    query.map(encodeURIComponent).join "+"

  # Create a search URL from the given :query (using either the provided search URL, or the default one).
  # It would be better to pull the default search engine from chrome itself.  However, chrome does not provide
  # an API for doing so.
  createSearchUrl: (query, searchUrl = Settings.get("searchUrl")) ->
    searchUrl += "%s" if -1 == searchUrl.indexOf("%s") + searchUrl.indexOf("%S")
    searchUrl = searchUrl.replace /%S/g, query
    searchUrl.replace /%s/g, @createSearchQuery query

  # Extract a query from url if it appears to be a URL created from the given search URL.
  # For example, map "https://www.google.ie/search?q=star+wars&foo&bar" to "star wars".
  extractQuery: do =>
    queryTerminator = new RegExp "[?&#/]"
    httpProtocolRegexp = new RegExp "^https?://"
    (searchUrl, url) ->
      url = url.replace httpProtocolRegexp
      searchUrl = searchUrl.replace httpProtocolRegexp
      [ searchUrl, suffixTerms... ] = searchUrl.split "%s"
      # We require the URL to start with the search URL.
      return null unless url.startsWith searchUrl
      # We require any remaining terms in the search URL to also be present in the URL.
      for suffix in suffixTerms
        return null unless 0 <= url.indexOf suffix
      # We use try/catch because decodeURIComponent can throw an exception.
      try
          url[searchUrl.length..].split(queryTerminator)[0].split("+").map(decodeURIComponent).join " "
      catch
        null

  # Converts :string into a Google search if it's not already a URL. We don't bother with escaping characters
  # as Chrome will do that for us.
  convertToUrl: (string) ->
    string = string.trim()

    # Special-case about:[url], view-source:[url] and the like
    if Utils.hasChromePrefix string
      string
    else if Utils.hasJavascriptPrefix string
      # In Chrome versions older than 46.0.2467.2, encoded javascript URIs weren't handled correctly.
      if Utils.haveChromeVersion "46.0.2467.2" then string else Utils.decodeURIByParts string
    else if Utils.isUrl string
      Utils.createFullUrl string
    else
      Utils.createSearchUrl string

  # detects both literals and dynamically created strings
  isString: (obj) -> typeof obj == 'string' or obj instanceof String

  # Transform "zjkjkabz" into "abjkz".
  distinctCharacters: (str) ->
    chars = str.split("").sort()
    (ch for ch, index in chars when index == 0 or ch != chars[index-1]).join ""

  # Compares two version strings (e.g. "1.1" and "1.5") and returns
  # -1 if versionA is < versionB, 0 if they're equal, and 1 if versionA is > versionB.
  compareVersions: (versionA, versionB) ->
    versionA = versionA.split(".")
    versionB = versionB.split(".")
    for i in [0...(Math.max(versionA.length, versionB.length))]
      a = parseInt(versionA[i] || 0, 10)
      b = parseInt(versionB[i] || 0, 10)
      if (a < b)
        return -1
      else if (a > b)
        return 1
    0

  # True if the current Chrome version is at least the required version.
  haveChromeVersion: (required) ->
    chromeVersion = navigator.appVersion.match(/Chrom(e|ium)\/(.*?) /)?[2]
    chromeVersion and 0 <= Utils.compareVersions chromeVersion, required

  # Zip two (or more) arrays:
  #   - Utils.zip([ [a,b], [1,2] ]) returns [ [a,1], [b,2] ]
  #   - Length of result is `arrays[0].length`.
  #   - Adapted from: http://stackoverflow.com/questions/4856717/javascript-equivalent-of-pythons-zip-function
  zip: (arrays) ->
    arrays[0].map (_,i) ->
      arrays.map( (array) -> array[i] )

  # locale-sensitive uppercase detection
  hasUpperCase: (s) -> s.toLowerCase() != s

  # Does string match any of these regexps?
  matchesAnyRegexp: (regexps, string) ->
    for re in regexps
      return true if re.test string
    false

  # Convenience wrapper for setTimeout (with the arguments around the other way).
  setTimeout: (ms, func) -> setTimeout func, ms

  # Like Nodejs's nextTick.
  nextTick: (func) -> @setTimeout 0, func

  # Make an idempotent function.
  makeIdempotent: (func) ->
    (args...) -> ([previousFunc, func] = [func, null])[0]? args...

  monitorChromeStorage: (key, setter) ->
    # NOTE: "?" here for the tests.
    chrome?.storage.local.get key, (obj) =>
      setter obj[key] if obj[key]?
      chrome.storage.onChanged.addListener (changes, area) =>
        setter changes[key].newValue if changes[key]?.newValue?

# This creates a new function out of an existing function, where the new function takes fewer arguments. This
# allows us to pass around functions instead of functions + a partial list of arguments.
Function::curry = ->
  fixedArguments = Array.copy(arguments)
  fn = this
  -> fn.apply(this, fixedArguments.concat(Array.copy(arguments)))

Array.copy = (array) -> Array.prototype.slice.call(array, 0)

String::startsWith = (str) -> @indexOf(str) == 0
String::ltrim = -> @replace /^\s+/, ""
String::rtrim = -> @replace /\s+$/, ""
String::reverse = -> @split("").reverse().join ""

globalRoot = window ? global
globalRoot.extend = (hash1, hash2) ->
  for own key of hash2
    hash1[key] = hash2[key]
  hash1

# A simple cache. Entries used within two expiry periods are retained, otherwise they are discarded.
# At most 2 * @entries entries are retained.
class SimpleCache
  # expiry: expiry time in milliseconds (default, one hour)
  # entries: maximum number of entries in @cache (there may be up to this many entries in @previous, too)
  constructor: (@expiry = 60 * 60 * 1000, @entries = 1000) ->
    @cache = {}
    @previous = {}
    @lastRotation = new Date()

  has: (key) ->
    @rotate()
    (key of @cache) or key of @previous

  # Set value, and return that value.  If value is null, then delete key.
  set: (key, value = null) ->
    @rotate()
    delete @previous[key]
    if value?
      @cache[key] = value
    else
      delete @cache[key]
      null

  get: (key) ->
    @rotate()
    if key of @cache
      @cache[key]
    else if key of @previous
      @cache[key] = @previous[key]
      delete @previous[key]
      @cache[key]
    else
      null

  rotate: (force = false) ->
    Utils.nextTick =>
      if force or @entries < Object.keys(@cache).length or @expiry < new Date() - @lastRotation
        @lastRotation = new Date()
        @previous = @cache
        @cache = {}

  clear: ->
    @rotate true
    @rotate true

# This is a simple class for the common case where we want to use some data value which may be immediately
# available, or for which we may have to wait.  It implements a use-immediately-or-wait queue, and calls the
# fetch function to fetch the data asynchronously.
class AsyncDataFetcher
  constructor: (fetch) ->
    @data = null
    @queue = []
    Utils.nextTick =>
      fetch (@data) =>
        callback @data for callback in @queue
        @queue = null

  use: (callback) ->
    if @data? then callback @data else @queue.push callback

# This takes a list of jobs (functions) and runs them, asynchronously.  Functions queued with @onReady() are
# run once all of the jobs have completed.
class JobRunner
  constructor: (@jobs) ->
    @fetcher = new AsyncDataFetcher (callback) =>
      for job in @jobs
        do (job) =>
          Utils.nextTick =>
            job =>
              @jobs = @jobs.filter (j) -> j != job
              callback true if @jobs.length == 0

  onReady: (callback) ->
    @fetcher.use callback

root = exports ? (window.root ?= {})
root.Utils = Utils
root.SimpleCache = SimpleCache
root.AsyncDataFetcher = AsyncDataFetcher
root.JobRunner = JobRunner
unless exports?
  root.extend = extend
  extend window, root