[GitHub] [daffodil] stevedlawrence commented on a diff in pull request #908: Refactor and document SAX unparse implementation

[GitBox <gi...@apache.org>]

stevedlawrence commented on code in PR #908:
URL: https://github.com/apache/daffodil/pull/908#discussion_r1067285755


##########
daffodil-runtime1/src/main/scala/org/apache/daffodil/infoset/SAXInfosetInputter.scala:
##########
@@ -21,79 +21,70 @@ import java.net.URI
 import java.net.URISyntaxException
 
 import org.apache.daffodil.api.DFDL
-import org.apache.daffodil.api.DFDL.DaffodilUnhandledSAXException
-import org.apache.daffodil.api.DFDL.DaffodilUnparseErrorSAXException
-import org.apache.daffodil.api.DFDL.SAXInfosetEvent
 import org.apache.daffodil.dpath.NodeInfo
 import org.apache.daffodil.exceptions.Assert
 import org.apache.daffodil.infoset.InfosetInputterEventType.EndDocument
 import org.apache.daffodil.infoset.InfosetInputterEventType.StartElement
+import org.apache.daffodil.processors.DaffodilUnparseContentHandler
+import org.apache.daffodil.util.Coroutine
+import org.apache.daffodil.util.Maybe
 import org.apache.daffodil.util.Maybe.One
+import org.apache.daffodil.util.Maybe.Nope
 import org.apache.daffodil.util.MaybeBoolean
 import org.apache.daffodil.util.Misc
 import org.apache.daffodil.xml.XMLUtils
 
 /**
- * The SAXInfosetInputter consumes SAXInfosetEvent objects from the DaffodilUnparseContentHandler
- * class and converts them to events that the DataProcessor unparse can use. This class contains an
- * array of batched SAXInfosetEvent objects that it receives from the contentHandler and the index
- * of the current element being processed.
+ * The SAXInfosetInputter consumes SAXInfosetEvent objects from the
+ * DaffodilUnparseContentHandler class and converts them to events that the
+ * DataProcessor unparse can use.
+ * 
+ * See the DaffodilUnparseContentHandler for a detailed description of how these two
+ * classes interact.
  *
- * This class, together with the SAXInfosetInputter, uses coroutines to ensure that a batch of events
- * (based on the tunable saxUnparseEventBatchSize) can be passed from the former to the latter.
- * The following is the general process:
- *
- * - the run method is called, with the first batch of events, starting with the StartDocument event,
- * already loaded on the inputter's queue.
- * This is collected and stored in the batchedInfosetEvents member, and the currentIndex is set to 0
- * - The dp.unparse method is called, and it calls hasNext to make sure an event exists to be
- * processed and then queries the event at currentIndex. The hasNext call also checks that there is
- * a next event to be processed (currentIndex+1), and if not, queues the next batch of events by
- * transferring control to the contentHandler so it can load them.
- * - After it is done with the current event, it calls inputter.next to get the next event, and that
- * increments the currentIndex and cleans out the event at the previous index
- * - This process continues until the event at currentIndex either contains an EndDocument event or
- * the currentIndex is the last in the batch. If it is the former, the endDocumentReceived flag is
- * set to true and hasNext will return false. If it is the latter, the next batch of events will be
- * queued by transferring control to the contentHandler so it can load them.
- * - This ends the unparse process, and the unparseResult and/or any Errors are set on a single element
- * array containing response events. We call resumeFinal passing along that array, terminating this
- * thread and resuming the contentHandler for the last time.
- *
- * @param unparseContentHandler producer coroutine that sends the SAXInfosetEvent to this class
- * @param dp dataprocessor that we use to kickstart the unparse process and that consumes the
- *           currentEvent
- * @param output  outputChannel of choice where the unparsed data is stored
+ * @param unparseContentHandler producer coroutine that sends the SAXInfosetEvents to this class
+ * @param dp DataProcessor that we use to kickstart the unparse process to consumes the events
+ * @param output  OutputChannel where the unparsed data is written
+ * @param resolveRelaiveInfosetBlobURIs if true, elements with type xs:anyURI type and a
+ *   relativeURI are resolved relative to the classpath. This should only be true when used
+ *   via a TDMLRunner or similar testing infrastructure
  */
 class SAXInfosetInputter(
-  unparseContentHandler: DFDL.DaffodilUnparseContentHandler,
+  unparseContentHandler: DaffodilUnparseContentHandler,
   dp: DFDL.DataProcessor,
-  output: DFDL.Output)
-  extends InfosetInputter with DFDL.ConsumerCoroutine {
+  output: DFDL.Output,
+  resolveRelativeInfosetBlobURIs: Boolean)
+  extends InfosetInputter with Coroutine[Array[SAXInfosetEvent]] {
 
   /**
-   * allows support for converting relative URIs in Blobs to absolute URIs, this is necessary
-   * for running TDML tests as they allow relative URIs. Because Daffodil proper only uses
-   * absolute URIs, we hide this functionality behind this flag. It can be set to true by calling
-   * the unparseContentHandler.enableInputterResolutionOfRelativeInfosetBlobURIs(), which calls the
-   * inputter's enableResolutionOfRelativeInfosetBlobURIs() function to set the below variable to true
+   * The index into batchedInfosetEvents that the InfosetInputter is currently returning
+   * information about to the unparse()
    */
-  private var resolveRelativeInfosetBlobURIs: Boolean = false
-
-  private var endDocumentReceived = false
   private var currentIndex: Int = 0
+
+  /**
+   * The array of batched infoset events received from the ContentHandler.
+   */

Review Comment:
   Yep, there would be one `resume()` call passing the entire batch to the infoset inputter coroutine, and one `resumeFinal()` call passing the `UnparseResult` back to the ContentHandler. Unfortunately there's still a thread created so there is that same Thread overhead.
   
   In theory, if we found the EndDocument in the first batch we could potentially even avoid the coroutine/thread stuff altogether. Since the SAXInofsetINputter threated wouldn't have been started yet we could have a setter to directly set the batch, and then call the unparse() function passing in that inputter. Could potentially be a big win for small-ish infosets. But does add extra complication. 



-- 
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.

To unsubscribe, e-mail: commits-unsubscribe@daffodil.apache.org

For queries about this service, please contact Infrastructure at:
users@infra.apache.org