Logo Questions Linux Laravel Mysql Ubuntu Git Menu
 

How does ng-repeat work?

I dissected ng-repeat and extracted the code blocks attached, seeing that these comprise the logic that handles the repeating algorithm (which I want to understand how it works).

I have quite a few questions, but since they are all about the internals of ng-repeat I chose to ask them all here. I don't see any reason to separate them into different SO questions. I have marked inline to which line(s) of code each question refers to.

  1. Why do they need to make sure that trackById is not the native hasOwnProperty function? (that's what that assertNotHasOwnProperty function does, part of Angular's internal API)
  2. As far as my intuition go, this code executes on items already in the repeater, when it has to update the collection - it just picks them up and pushes them into the list for processing, right?
  3. This code block obviously looks for duplicates in the repeater collection. But how exactly does it do that is beyond me. Please explain.
  4. Why does Angular have to store the block object both nextBlockMap and in nextBlockOrder?
  5. What are block.endNode and block.startNode?
  6. I assume the answer to the above question will clarify how this algorithm work, but please explain why it has to check if the nextNode has (been) '$$NG_REMOVED'?
  7. What happens here? Again, I assume question 6 will already provide an answer to this one. But still pointing that out.

Like I said, I dug through ng-repeat to find the code I deemed relevant to the repeating mechanism. Plus, I do understand the rest of the directive. So without further ado, here is the code (from v1.2.0):

      length = nextBlockOrder.length = collectionKeys.length;
      for (index = 0; index < length; index++) {
       key = (collection === collectionKeys) ? index : collectionKeys[index];
       value = collection[key];
       trackById = trackByIdFn(key, value, index);

       // question #1
       assertNotHasOwnProperty(trackById, '`track by` id');

       // question #2
       if (lastBlockMap.hasOwnProperty(trackById)) {
         block = lastBlockMap[trackById];
         delete lastBlockMap[trackById];
         nextBlockMap[trackById] = block;
         nextBlockOrder[index] = block;

       // question #3
       } else if (nextBlockMap.hasOwnProperty(trackById)) {
         // restore lastBlockMap
         forEach(nextBlockOrder, function(block) {
           if (block && block.startNode) lastBlockMap[block.id] = block;
         });
         // This is a duplicate and we need to throw an error
         throw ngRepeatMinErr('dupes', "Duplicates in a repeater are not allowed. Use 'track by' expression to specify unique keys. Repeater: {0}, Duplicate key: {1}",
                                                                                                                                                expression,       trackById);

       // question #4
       } else {
         // new never before seen block
         nextBlockOrder[index] = { id: trackById };
         nextBlockMap[trackById] = false;
       }
     }


      for (index = 0, length = collectionKeys.length; index < length; index++) {
        key = (collection === collectionKeys) ? index : collectionKeys[index];
        value = collection[key];
        block = nextBlockOrder[index];


        // question #5
        if (nextBlockOrder[index - 1]) previousNode = nextBlockOrder[index - 1].endNode;

        if (block.startNode) {
          // if we have already seen this object, then we need to reuse the
          // associated scope/element
          childScope = block.scope;

          // question #6
          nextNode = previousNode;
          do {
            nextNode = nextNode.nextSibling;
          } while(nextNode && nextNode[NG_REMOVED]);
          if (block.startNode != nextNode) {
            // existing item which got moved
            $animate.move(getBlockElements(block), null, jqLite(previousNode));
          }
          previousNode = block.endNode;

        } else {
          // new item which we don't know about
          childScope = $scope.$new();
        }

        // question #7
        if (!block.startNode) {
          linker(childScope, function(clone) {
            clone[clone.length++] = document.createComment(' end ngRepeat: ' + expression + ' ');
            $animate.enter(clone, null, jqLite(previousNode));
            previousNode = clone;
            block.scope = childScope;
            block.startNode = previousNode && previousNode.endNode ? previousNode.endNode : clone[0];
            block.endNode = clone[clone.length - 1];
            nextBlockMap[block.id] = block;
          });
        }
      }
      lastBlockMap = nextBlockMap;
like image 437
pilau Avatar asked Nov 21 '13 22:11

pilau


1 Answers

After some tinkering with the directive, I became familiar with ng-repeaters code, and managed to answer some of my questions. I highlighted in bold the things I couldn't yet figure out on my own, and would appreciate if anyone could shed some light on the bold parts:

  1. The ID is tested for hasOwnProperty, because they use that method to check whether the ID is present in the iteration objects (lastBlockMap, nextBlockMap) (this process explained below). I couldn't find out on what scenario this can actually happen, however.
  2. I was correct in my assumption. nextBlockMap contains all items that will be transcluded on the current model change. lastBlockMap contains everything from the previous model update. It used for finding duplicates in the collection.
  3. OK, this one is pretty straightforward actually. In this for loop, ng-repeat fills up nextBlockMap with items from lastBlockMap. Looking at the order of ifs, it's easy to see that if the item cannot be found in lastBlockMap, but it is already present in nextBlockMap (meaning, it was already copied there from lastBlockMap, and therefore its trackById appears twice in the collection) - it's a duplicate. What the forEach does is simply run through all initialized items in nextBlockMap (blocks that have a startNode property) and push their ID back into lastBlockMap. I cannot however understand why this is necessary.
  4. The only reason I could find for separating nextBlockOrder (all trackByIds in an array) from nextBlockMap (all block objects in a trackById hash), is this line, which working with an array makes it an easy and simple operation: if (nextBlockOrder[index - 1]) previousNode = nextBlockOrder[index - 1].endNode;. It is explained in the answers to question 5 and 6:
  5. block.startNode and block.endNode are the first and last DOM nodes in the block belonging to an item in the collected being repeated. Therefore, this line here sets previousNode to reference the last DOM node of the previous item in the repeater.
  6. previousNode is then used as the first node, in a loop that checks how the DOM changed when items have been moved around or removed from the repeater collection - again, only in case we are not working with the first block in the array.
  7. This is easy - it initializes the block - assigning the $scope and startNode and endNode for later reference, and saves everything in nextBlockMap. The comment created right after the cloned element, is there to guarantee we always have an endNode.
like image 71
pilau Avatar answered Sep 20 '22 23:09

pilau