Tidy HTML before trying to convert it with abiword

settings.json.template

@@ -90,6 +90,10 @@
      Abiword is needed to advanced import/export features of pads*/
   "abiword" : null,
 
+  /* This is the path to the Tidy executable. Setting it to null, disables Tidy.
+     Tidy is used to improve the quality of exported pads*/
+  "tidyHtml" : null,
+
   /* Allow import of file types other than the supported types: txt, doc, docx, rtf, odt, html & htm */
   "allowUnknownFileEnds" : true,
 

src/node/handler/ExportHandler.js

@@ -28,6 +28,7 @@ var fs = require("fs");
 var settings = require('../utils/Settings');
 var os = require('os');
 var hooks = require("ep_etherpad-lite/static/js/pluginfw/hooks");
+var TidyHtml = require('../utils/TidyHtml');
 
 //load abiword only if its enabled
 if(settings.abiword != null)
@@ -172,12 +173,19 @@ exports.doExport = function(req, res, padId, type)
               fs.writeFile(srcFile, html, callback);
             }
           },
-          //send the convert job to abiword
+
+          // Tidy up the exported HTML
           function(callback)
           {
             //ensure html can be collected by the garbage collector
             html = null;
 
+            TidyHtml.tidy(srcFile, callback);
+          },
+
+          //send the convert job to abiword
+          function(callback)
+          {
             destFile = tempDirectory + "/etherpad_export_" + randNum + "." + type;
             abiword.convertFile(srcFile, destFile, type, callback);
           },

src/node/utils/Settings.js

@@ -152,6 +152,11 @@ exports.minify = true;
  */
 exports.abiword = null;
 
+/**
+ * The path of the tidy executable
+ */
+exports.tidyHtml = null;
+
 /**
  * Should we support none natively supported file types on import?
  */

src/node/utils/TidyHtml.js

@@ -0,0 +1,35 @@
+/**
+ * Tidy up the HTML in a given file
+ */
+
+var settings = require("./Settings");
+var spawn = require('child_process').spawn;
+
+exports.tidy = function(srcFile, callback) {
+  // Don't do anything if Tidy hasn't been enabled
+  if (!settings.tidyHtml) {
+    return callback(null);
+  }
+
+  var errMessage = '';
+
+  // Spawn a new tidy instance that cleans up the file inline
+  var tidy = spawn(settings.tidyHtml, ['-modify', srcFile]);
+
+  // Keep track of any error messages
+  tidy.stderr.on('data', function (data) {
+    errMessage += data.toString();
+  });
+
+  // Wait until Tidy is done
+  tidy.on('close', function(code) {
+    // Tidy returns a 0 when no errors occur and a 1 exit code when
+    // the file could be tidied but a few warnings were generated
+    if (code === 0 || code === 1) {
+      return callback(null);
+    } else {
+      console.error(errMessage);
+      return callback('Tidy died with exit code ' + code);
+    }
+  });
+};