XMLEventReader OutOfMemoryError on very long but flat XML input

[scabug]

XMLEventReader can run out of memory on very long input of this type:

<foo>
  <bar/>
  <bar/>
  <!-- millions more <bar/> -->
</foo>

This type of input is typical for a table export with lots of rows.

I've traced this back to XMLEventReader.elem which generates dummy tags. These tags are then collected into the parent node in a NodeBuffer. Eventually, the NodeBuffer exhausts the available heap.

[scabug]

Imported From: https://issues.scala-lang.org/browse/SI-3840?orig=1
Reporter: huynhjl
Attachments:

• 0001-Modified-XMLEventReader-to-write-dummy-ignore-tag-on.patch (created on Sep 14, 2010 12:46:46 PM UTC, 2456 bytes)
• ManyChildrenOOM.scala (created on Sep 14, 2010 12:44:48 PM UTC, 1018 bytes)

[scabug]

huynhjl said:
Sample object file that causes the behavior

[scabug]

huynhjl said:
Patch to XMLEventReader (returns NodeSeq.Empty except for top node)

[scabug]

@paulp said:
(In r23018) Modified XMLEventReader to write dummy tag only once.
Contributed by Jean-Laurent Huynh, reviewed by extempore. Closes #3840.