Merge pull request #6 from legalthings/upgrade-instructions

Upgrade instructions

README.md

@@ -41,3 +41,233 @@ Include `viewer.html` using [SSI](http://httpd.apache.org/docs/2.4/howto/ssi.htm
 </html>
 ```
 
+## Upgrading the source
+
+Normally mozilla's PDF js viewer, will only run as standalone. This version is patched so you can include it within a
+page.
+
+To update this version, get the pdf.js source code and build the project
+
+    git clone https://github.com/mozilla/pdf.js.git
+    cd pdf.js
+    npm install
+    node make generic
+    cd ..
+
+And update the files from source and patch them
+
+    cd pdf.js-viewer
+    npm install
+    ./build.sh ../pdf.js/build/generic/
+
+### Manual patching
+
+When updating to a new minor (or major) version, it's likely than one or more of the chunks can't be applied. This
+means you need to do these modifications manually.
+
+
+#### function getL10nData()
+
+The viewer uses `l10n.js` with a `<link rel="resource" type="application/l10n">` header for internationalization. This
+chunk makes using that optional.
+
+       function getL10nData(key, args, fallback) {
+         var data = gL10nData[key];
+         if (!data) {
+    -      console.warn('#' + key + ' is undefined.');
+    +      if (Object.keys(gL10nData).length > 0) {
+    +        console.warn('#' + key + ' is undefined.');
+    +      }
+           if (!fallback) {
+             return null;
+           }
+
+#### Dynamic paths
+
+The viewer uses relative paths to JavaScript files. This doesn't work when the viewer is embedded on a web page.
+Instead the paths are determined based on the path of the current JavaScript file.
+
+    -PDFJS.imageResourcesPath = './images/';
+    -  PDFJS.workerSrc = '../build/pdf.worker.js';
+    -  PDFJS.cMapUrl = '../web/cmaps/';
+    -  PDFJS.cMapPacked = true;
+    +var scriptTagContainer = document.body ||
+    +                         document.getElementsByTagName('head')[0];
+    +var pdfjsSrc = scriptTagContainer.lastChild.src;
+    +
+    +if (pdfjsSrc) {
+    +  PDFJS.imageResourcesPath = pdfjsSrc.replace(/pdf\.js$/i, 'images/');
+    +  PDFJS.workerSrc = pdfjsSrc.replace(/pdf\.js$/i, 'pdf.worker.js');
+    +  PDFJS.cMapUrl = pdfjsSrc.replace(/pdf\.js$/i, 'cmaps/');
+    +}
+    +
+    +PDFJS.cMapPacked = true;
+
+#### Explicitly load a PDF document
+
+The viewer shouldn't start loading a (default) document when it's loaded. Instead we want to expose the initialization,
+so it can be called in JavaScript with `PDFJS.webViewerLoad()`.
+
+    -document.addEventListener('DOMContentLoaded', webViewerLoad, true);
+    +// document.addEventListener('DOMContentLoaded', webViewerLoad, true);
+    +PDFJS.webViewerLoad = function (src) {
+    +  if (src) DEFAULT_URL = src;
+    +
+    +  webViewerLoad();
+    +}
+
+On several places the code assumes that a PDF is loaded, which (because of the explicit load) might not be the case. We
+need to check if `pdfDocument` is set before using it.
+
+##### PDFViewerApplication.pagesCount() and PDFLinkService.pagesCount()
+
+         get pagesCount() {
+    -      return this.pdfDocument.numPages;
+    +      return this.pdfDocument ? this.pdfDocument.numPages : 0;
+         },
+
+_The pagesCount method for both `PDFViewerApplication` and `PDFLinkService`. Both need to be patched._
+
+##### PDFViewerApplication.cleanup()
+
+       cleanup: function pdfViewCleanup() {
+         this.pdfViewer.cleanup();
+         this.pdfThumbnailViewer.cleanup();
+    -    this.pdfDocument.cleanup();
+    +    if (this.pdfDocument) {
+    +      this.pdfDocument.cleanup();
+    +    }
+       },
+
+#### overlayManagerRegister
+
+The overlay is registered when the viewer is loaded. The original code will only do this once and give an error on each
+subsequent call. Escpecially with single-page applications (eg an Angular app), the viewer may be loaded multiple times.
+This patch causes pdf.js to unregister an unusued (closed) overlay.
+
+       register: function overlayManagerRegister(name, callerCloseMethod, canForceClose) {
+         return new Promise(function (resolve) {
+           var element, container;
+           if (!name || !(element = document.getElementById(name)) ||
+               !(container = element.parentNode)) {
+             throw new Error('Not enough parameters.');
+           } else if (this.overlays[name]) {
+    -        throw new Error('The overlay is already registered.');
+    +        if (this.active !== name) {
+    +          this.unregister(name);
+    +        } else {
+    +          throw new Error('The overlay is already registered and active.');
+    +        }
+           }
+
+#### webViewerChange trigger on file upload dialog
+
+Whenever a file dialog is used (so with any `<input type="file">` on the page), the file is loaded into the pdf.js
+viewer. We don't want this behaviour, so comment it out.
+
+    -window.addEventListener('change', function webViewerChange(evt) {
+    +/*window.addEventListener('change', function webViewerChange(evt) {
+       var files = evt.target.files;
+       if (!files || files.length === 0) {
+         return;
+    @@ -17627,7 +17647,7 @@
+         setAttribute('hidden', 'true');
+       document.getElementById('download').setAttribute('hidden', 'true');
+       document.getElementById('secondaryDownload').setAttribute('hidden', 'true');
+    -}, true);
+    +}, true);*/
+
+#### handleMouseWheel()
+
+The JavaScript pdf.js file might be loaded while the viewer isn't being displayed. This causes an error on mouse move.
+We need to check if the viewer is initialized, before handling the event.
+
+     function handleMouseWheel(evt) {
+    +  // Ignore mousewheel event if pdfViewer isn't loaded
+    +  if (!PDFViewerApplication.pdfViewer) return;
+
+#### Load code for worker using AJAX if needed
+
+A [Web Worker](https://developer.mozilla.org/en-US/docs/Web/API/Web_Workers_API/Using_web_workers) can't use code from
+a same-origin domain. The CORS headers don't apply.
+
+he patch will cause pdf.js to first try to create the Worker the regular way, with a URL to the JavaScript source. If
+this fails, the source if fetched using AJAX and used to create an
+[object url](https://developer.mozilla.org/en-US/docs/Web/API/URL/createObjectURL). If this also fails, pdf.js will go
+onto it's last resort by calling `setupFakeWorker()`.
+
+    +    /**
+    +     * Needed because workers cannot load scripts outside of the current origin (as of firefox v45).
+    +     * This patch does require the worker script to be served with a (Access-Control-Allow-Origin: *) header
+    +     * @patch
+    +     */
+    +    var loadWorkerXHR = function(){
+    +      var url = PDFJS.workerSrc;
+    +      var jsdfd = PDFJS.createPromiseCapability();
+    +
+    +      if (url.match(/^blob:/) || typeof URL.createObjectURL === 'undefined') {
+    +        jsdfd.reject(); // Failed loading using blob
+    +      }
+    +
+    +      var xmlhttp;
+    +      xmlhttp = new XMLHttpRequest();
+    +
+    +      xmlhttp.onreadystatechange = function(){
+    +        if (xmlhttp.readyState != 4) return;
+    +
+    +        if (xmlhttp.status == 200) {
+    +          info('Loaded worker source through XHR.');
+    +          var workerJSBlob = new Blob([xmlhttp.responseText], { type: 'text/javascript' });
+    +          jsdfd.resolve(window.URL.createObjectURL(workerJSBlob));
+    +        } else {
+    +          jsdfd.reject();
+    +        }
+    +      };
+    +
+    +      xmlhttp.open('GET', url, true);
+    +      xmlhttp.send();
+    +      return jsdfd.promise;
+    +    }
+    +
+    +    var workerError = function() {
+    +      loadWorkerXHR().then(function(blob) {
+    +        PDFJS.workerSrc = blob;
+    +        loadWorker();
+    +      }, function() {
+    +        this.setupFakeWorker();
+    +      }.bind(this));
+    +    }.bind(this);
+    +
+
+    -    if (!globalScope.PDFJS.disableWorker && typeof Worker !== 'undefined') {
+    +    var loadWorker = function() {
+           var workerSrc = PDFJS.workerSrc;
+           if (!workerSrc) {
+             error('No PDFJS.workerSrc specified');
+    @@ -3559,6 +3603,8 @@
+             // Some versions of FF can't create a worker on localhost, see:
+             // https://bugzilla.mozilla.org/show_bug.cgi?id=683280
+             var worker = new Worker(workerSrc);
+    +        worker.onerror = workerError;
+    +
+             var messageHandler = new MessageHandler('main', worker);
+             this.messageHandler = messageHandler;
+
+    @@ -3589,11 +3635,16 @@
+             return;
+           } catch (e) {
+             info('The worker has been disabled.');
+    +        workerError();
+           }
+    -    }
+    +    }.bind(this);
+         // Either workers are disabled, not supported or have thrown an exception.
+         // Thus, we fallback to a faked worker.
+    -    this.setupFakeWorker();
+    +    if (!globalScope.PDFJS.disableWorker && typeof Worker !== 'undefined') {
+    +      loadWorker();
+    +    } else {
+    +      this.setupFakeWorker();
+    +    }
+       }
+

build.sh

@@ -1,14 +1,28 @@
-cat source/web/l10n.js source/build/pdf.js source/web/compatibility.js source/web/debugger.js source/web/viewer.js > pdf.js
+#!/bin/bash
+
+if [[ $# < 1 ]]; then
+  echo USAGE $0 SOURCE_PATH 1>&2
+  exit 1
+fi
+
+source="$1"
+
+if [[ ! -f "$source/build/pdf.js" ]]; then
+  echo "$source/build/pdf.js" not found 1>&2
+  exit 1
+fi
+
+cat "$source/web/l10n.js" "$source/build/pdf.js" "$source/web/compatibility.js" "$source/web/debugger.js" "$source/web/viewer.js" > pdf.js
 patch pdf.js < pdf.js.patch
 
-cp source/build/pdf.worker.js .
+cp "$source/build/pdf.worker.js" .
 patch pdf.worker.js < pdf.worker.js.patch
 
-uglifycss source/web/viewer.css > viewer.css
+uglifycss "$source/web/viewer.css" > viewer.css
 node grunt/css-prefix.js viewer.css viewer.css pdfjs
 
-sed -r 's/url\((")?images\//url\(\1@pdfjsImagePath\//g' < source/web/viewer.css | uglifycss > viewer.less
+sed -r 's/url\((")?images\//url\(\1@pdfjsImagePath\//g' < "$source/web/viewer.css" | uglifycss > viewer.less
 node grunt/css-prefix.js viewer.less viewer.less pdfjs
 
-cp source/web/cmaps/ source/web/images/ source/web/locale/ . -a
+cp "$source/web/cmaps/" "$source/web/images/" "$source/web/locale/" . -a