diff options
| author | Stephen Blott | 2014-11-08 07:40:49 +0000 |
|---|---|---|
| committer | Stephen Blott | 2014-11-08 07:40:49 +0000 |
| commit | 3d647c9d062a84082d0337ac0b0004d31eb64969 (patch) | |
| tree | 9ff193d27f26d931b06018e36662cde2faef2a91 | |
| parent | 8e4e95c87d2ea06fcc0d9d354a180dfa1f31cdd2 (diff) | |
| parent | 56565a021ae421d648a3b4180fdd255270245421 (diff) | |
| download | vimium-3d647c9d062a84082d0337ac0b0004d31eb64969.tar.bz2 | |
Merge branch 'smblott-github-domains-only-domains'
| -rw-r--r-- | background_scripts/completion.coffee | 8 | ||||
| -rw-r--r-- | lib/utils.coffee | 19 | ||||
| -rw-r--r-- | tests/unit_tests/completion_test.coffee | 12 | ||||
| -rw-r--r-- | tests/unit_tests/utils_test.coffee | 7 |
4 files changed, 28 insertions, 18 deletions
diff --git a/background_scripts/completion.coffee b/background_scripts/completion.coffee index 23696185..b75ebb87 100644 --- a/background_scripts/completion.coffee +++ b/background_scripts/completion.coffee @@ -238,7 +238,7 @@ class DomainCompleter onComplete() onPageVisited: (newPage) -> - domain = @parseDomain(newPage.url) + domain = @parseDomainAndScheme newPage.url if domain slot = @domains[domain] ||= { entry: newPage, referenceCount: 0 } # We want each entry in our domains hash to point to the most recent History entry for that domain. @@ -250,11 +250,13 @@ class DomainCompleter @domains = {} else toRemove.urls.forEach (url) => - domain = @parseDomain(url) + domain = @parseDomainAndScheme url if domain and @domains[domain] and ( @domains[domain].referenceCount -= 1 ) == 0 delete @domains[domain] - parseDomain: (url) -> url.split("/")[2] || "" + # Return something like "http://www.example.com" or false. + parseDomainAndScheme: (url) -> + Utils.hasFullUrlPrefix(url) and not Utils.hasChromePrefix(url) and url.split("/",3).join "/" # Suggestions from the Domain completer have the maximum relevancy. They should be shown first in the list. computeRelevancy: -> 1 diff --git a/lib/utils.coffee b/lib/utils.coffee index bbcee1a0..b7f8731a 100644 --- a/lib/utils.coffee +++ b/lib/utils.coffee @@ -26,28 +26,29 @@ Utils = -> id += 1 hasChromePrefix: do -> - chromePrefixes = [ "about:", "view-source:", "chrome-extension:", "data:" ] + chromePrefixes = [ "about:", "view-source:", "extension:", "chrome-extension:", "data:" ] (url) -> if 0 < url.indexOf ":" for prefix in chromePrefixes return true if url.startsWith prefix false + hasFullUrlPrefix: do -> + urlPrefix = new RegExp "^[a-z]{3,}://." + (url) -> urlPrefix.test url + # Completes a partial URL (without scheme) createFullUrl: (partialUrl) -> - unless /^[a-z]{3,}:\/\//.test partialUrl - "http://" + partialUrl - else - partialUrl + if @hasFullUrlPrefix(partialUrl) then partialUrl else ("http://" + partialUrl) # Tries to detect if :str is a valid URL. isUrl: (str) -> - # Starts with a scheme: URL - return true if /^[a-z]{3,}:\/\//.test 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 @@ -98,7 +99,7 @@ Utils = convertToUrl: (string) -> string = string.trim() - # Special-case about:[url] and view-source:[url] + # Special-case about:[url], view-source:[url] and the like if Utils.hasChromePrefix string string else if Utils.isUrl string diff --git a/tests/unit_tests/completion_test.coffee b/tests/unit_tests/completion_test.coffee index 811436a9..44989267 100644 --- a/tests/unit_tests/completion_test.coffee +++ b/tests/unit_tests/completion_test.coffee @@ -163,13 +163,13 @@ context "domain completer", should "return only a single matching domain", -> results = filterCompleter(@completer, ["story"]) - assert.arrayEqual ["history1.com"], results.map (result) -> result.url + assert.arrayEqual ["http://history1.com"], results.map (result) -> result.url should "pick domains which are more recent", -> # These domains are the same except for their last visited time. - assert.equal "history1.com", filterCompleter(@completer, ["story"])[0].url + assert.equal "http://history1.com", filterCompleter(@completer, ["story"])[0].url @history2.lastVisitTime = hours(3) - assert.equal "history2.com", filterCompleter(@completer, ["story"])[0].url + assert.equal "http://history2.com", filterCompleter(@completer, ["story"])[0].url should "returns no results when there's more than one query term, because clearly it's not a domain", -> assert.arrayEqual [], filterCompleter(@completer, ["his", "tory"]) @@ -194,15 +194,15 @@ context "domain completer (removing entries)", should "remove 1 entry for domain with reference count of 1", -> @onVisitRemovedListener { allHistory: false, urls: [@history1.url] } - assert.equal "history2.com", filterCompleter(@completer, ["story"])[0].url + assert.equal "http://history2.com", filterCompleter(@completer, ["story"])[0].url assert.equal 0, filterCompleter(@completer, ["story1"]).length should "remove 2 entries for domain with reference count of 2", -> @onVisitRemovedListener { allHistory: false, urls: [@history2.url] } - assert.equal "history2.com", filterCompleter(@completer, ["story2"])[0].url + assert.equal "http://history2.com", filterCompleter(@completer, ["story2"])[0].url @onVisitRemovedListener { allHistory: false, urls: [@history3.url] } assert.equal 0, filterCompleter(@completer, ["story2"]).length - assert.equal "history1.com", filterCompleter(@completer, ["story"])[0].url + assert.equal "http://history1.com", filterCompleter(@completer, ["story"])[0].url should "remove 3 (all) matching domain entries", -> @onVisitRemovedListener { allHistory: false, urls: [@history2.url] } diff --git a/tests/unit_tests/utils_test.coffee b/tests/unit_tests/utils_test.coffee index b2d656ab..556f5b7a 100644 --- a/tests/unit_tests/utils_test.coffee +++ b/tests/unit_tests/utils_test.coffee @@ -61,6 +61,13 @@ context "hasChromePrefix", assert.isFalse Utils.hasChromePrefix "data" assert.isFalse Utils.hasChromePrefix "data :foobar" +context "isUrl", + should "identify URLs as URLs", -> + assert.isTrue Utils.isUrl "http://www.example.com/blah" + + should "identify non-URLs and non-URLs", -> + assert.isFalse Utils.isUrl "http://www.example.com/ blah" + context "Function currying", should "Curry correctly", -> foo = (a, b) -> "#{a},#{b}" |
