From 21fdaa4f2d851c8fa8c29c4a7a087041a2a4121d Mon Sep 17 00:00:00 2001
From: Matthew Lyon <matthew@lyonheart.us>
Date: Fri, 7 Mar 2014 11:57:12 -0800
Subject: [PATCH 1/4] widont: Add failing tests for short texts

---
 test/typogr.test.js | 5 +++++
 1 file changed, 5 insertions(+)

diff --git a/test/typogr.test.js b/test/typogr.test.js
index 843186d..cb284fb 100644
--- a/test/typogr.test.js
+++ b/test/typogr.test.js
@@ -52,6 +52,11 @@ module.exports = {
     assert.equal(tp.widont(' Test'), ' Test');
     assert.equal(tp.widont('<ul><li>Test</p></li><ul>'), '<ul><li>Test</p></li><ul>');
     assert.equal(tp.widont('<ul><li> Test</p></li><ul>'), '<ul><li> Test</p></li><ul>');
+    // neither should two or three word items
+    assert.equal(tp.widont('<h1>One Two</h1>'), '<h1>One Two</h1>')
+    assert.equal(tp.widont('<h1>One Two Three</h1>'), '<h1>One Two Three</h1>')
+    assert.equal(tp.widont('<h1>One  Two  Three</h1>'), '<h1>One  Two  Three</h1>')
+
     assert.equal(tp.widont('<p>In a couple of paragraphs</p><p>paragraph two</p>'),
                            '<p>In a couple of&nbsp;paragraphs</p><p>paragraph&nbsp;two</p>');
     assert.equal(tp.widont('<h1><a href="#">In a link inside a heading</i> </a></h1>'),

From f43559a0e167d072d590ecb11a3d313c07861d6d Mon Sep 17 00:00:00 2001
From: Matthew Lyon <matthew@lyonheart.us>
Date: Fri, 7 Mar 2014 11:57:50 -0800
Subject: [PATCH 2/4] widont: Pass failing tests for short texts

This perhaps isn't as good of a solution as it could be since it doesn't tie
into the existing regexp, but it is simple and doesn't trample on any of the
other tests.  It also allows me to provide awareness of the problem and provide
a proposed solution, without investing too much time in trying to modify the
regexp.
---
 typogr.js | 1 +
 1 file changed, 1 insertion(+)

diff --git a/typogr.js b/typogr.js
index 634a097..83c82b2 100644
--- a/typogr.js
+++ b/typogr.js
@@ -134,6 +134,7 @@
    *
    */
   var widont = typogr.widont = function(text) {
+    if (text.split(/\s+/).length < 4) { return text; }             // don't bother with short texts.
     var re_widont = re(
           '([\\w,])'+                                              // a word boundry character or comma, matching group 1
           '(?:\\s+)'+                                              // one or more space characters

From 2b7d8908356092fe56ff26955cb1ccc8b5739605 Mon Sep 17 00:00:00 2001
From: Matthew Lyon <matthew@lyonheart.us>
Date: Fri, 14 Mar 2014 17:04:19 -0700
Subject: [PATCH 3/4] Leverage the regexp to handle word count

This will also handle strings like "<h1>One Two</h1><p>Three Four Five</p>" or
html tags with spaces in them for attributes.
---
 test/typogr.test.js | 4 ++--
 typogr.js           | 3 +--
 2 files changed, 3 insertions(+), 4 deletions(-)

diff --git a/test/typogr.test.js b/test/typogr.test.js
index cb284fb..4f3e182 100644
--- a/test/typogr.test.js
+++ b/test/typogr.test.js
@@ -57,8 +57,8 @@ module.exports = {
     assert.equal(tp.widont('<h1>One Two Three</h1>'), '<h1>One Two Three</h1>')
     assert.equal(tp.widont('<h1>One  Two  Three</h1>'), '<h1>One  Two  Three</h1>')
 
-    assert.equal(tp.widont('<p>In a couple of paragraphs</p><p>paragraph two</p>'),
-                           '<p>In a couple of&nbsp;paragraphs</p><p>paragraph&nbsp;two</p>');
+    assert.equal(tp.widont('<p>In a couple of paragraphs</p><p>the paragraph number two</p>'),
+                           '<p>In a couple of&nbsp;paragraphs</p><p>the paragraph number&nbsp;two</p>');
     assert.equal(tp.widont('<h1><a href="#">In a link inside a heading</i> </a></h1>'),
                            '<h1><a href="#">In a link inside a&nbsp;heading</i> </a></h1>');
     assert.equal(tp.widont('<h1><a href="#">In a link</a> followed by other text</h1>'),
diff --git a/typogr.js b/typogr.js
index 83c82b2..826501e 100644
--- a/typogr.js
+++ b/typogr.js
@@ -134,9 +134,8 @@
    *
    */
   var widont = typogr.widont = function(text) {
-    if (text.split(/\s+/).length < 4) { return text; }             // don't bother with short texts.
     var re_widont = re(
-          '([\\w,])'+                                              // a word boundry character or comma, matching group 1
+          '(\\s+[^\\s]+\\s+[^\\s]+)'+                              // matching group 1: two words preceeded by space, need four words to trigger
           '(?:\\s+)'+                                              // one or more space characters
           '('+                                                     // matching group 2
             '[^<>\\s]+'+                                           // nontag/nonspace characters

From b1411ef5286243b54f2bb2843efd45ab886d14f7 Mon Sep 17 00:00:00 2001
From: Matthew Lyon <matthew@lyonheart.us>
Date: Fri, 14 Mar 2014 17:41:39 -0700
Subject: [PATCH 4/4] Don't count inline tags as words

Thus, for "<h1>Foo <a href='#'>bar baz</a></h1>", the h1 tag has for the
purposes of the matching regexp three words.
---
 test/typogr.test.js | 1 +
 typogr.js           | 7 ++++++-
 2 files changed, 7 insertions(+), 1 deletion(-)

diff --git a/test/typogr.test.js b/test/typogr.test.js
index 4f3e182..96bfd0f 100644
--- a/test/typogr.test.js
+++ b/test/typogr.test.js
@@ -56,6 +56,7 @@ module.exports = {
     assert.equal(tp.widont('<h1>One Two</h1>'), '<h1>One Two</h1>')
     assert.equal(tp.widont('<h1>One Two Three</h1>'), '<h1>One Two Three</h1>')
     assert.equal(tp.widont('<h1>One  Two  Three</h1>'), '<h1>One  Two  Three</h1>')
+    assert.equal(tp.widont('<h1><a href="#">Links</a> should work</h1>'), '<h1><a href="#">Links</a> should work</h1>')
 
     assert.equal(tp.widont('<p>In a couple of paragraphs</p><p>the paragraph number two</p>'),
                            '<p>In a couple of&nbsp;paragraphs</p><p>the paragraph number&nbsp;two</p>');
diff --git a/typogr.js b/typogr.js
index 826501e..62a506f 100644
--- a/typogr.js
+++ b/typogr.js
@@ -134,8 +134,13 @@
    *
    */
   var widont = typogr.widont = function(text) {
+    var inline_tags = 'a|em|span|strong|i|b'
+    var word = '(?:<(?:'+inline_tags+')[^>]*?>)*?[^\\s<>]+(?:</(?:'+inline_tags+')[^>]*?>)*?'
     var re_widont = re(
-          '(\\s+[^\\s]+\\s+[^\\s]+)'+                              // matching group 1: two words preceeded by space, need four words to trigger
+          '('+                                                     // matching group 1
+            '\\s+'+word+                                           // space and a word with a possible bordering tag
+            '\\s+'+word+                                           // space and a word with a possible bordering tag
+          ')'+
           '(?:\\s+)'+                                              // one or more space characters
           '('+                                                     // matching group 2
             '[^<>\\s]+'+                                           // nontag/nonspace characters