Jonathan Prates created PDFBOX-5824:
---------------------------------------
Summary: Allow COSDictionary.MAP_THRESHOLD to be defined as System
property
Key: PDFBOX-5824
URL: https://issues.apache.org/jira/browse/PDFBOX-5824
Project: PDFBox
Issue Type: Improvement
Components: PDModel
Affects Versions: 3.0.3 PDFBox, 4.0.0
Reporter: Jonathan Prates
Attachments: Screenshot 2024-05-21 at 11.00.25.png
[COSDictionary.MAP_THRESHOLD|https://github.com/apache/pdfbox/blob/trunk/pdfbox/src/main/java/org/apache/pdfbox/cos/COSDictionary.java#L54]
controls which Map class is used to optimize memory usage. By default, a
SmallMap is used. However, if the number of items in a COSDictionary reaches
the MAP_THRESHOLD value (hardcoded to 1,000), the references [are copied
|https://github.com/apache/pdfbox/blob/trunk/pdfbox/src/main/java/org/apache/pdfbox/cos/COSDictionary.java#L208]to
a LinkedHashMap.
For larger documents, where the COSDictionary is expected to be substantial,
this copying occurs frequently. Additionally, [SmallMap.keySet is not
efficient|https://github.com/apache/pdfbox/blob/trunk/pdfbox/src/main/java/org/apache/pdfbox/util/SmallMap.java#L281].
The attached screenshot shows pdfbox performance with SmallMap (in red) versus
using LinkedHashMap and ignoring the threshold (in green).
*Would it be beneficial to allow MAP_THRESHOLD to be defined as a System
property?*
If set to 0, LinkedHashMap would be used. If not set, it would default to the
current MAP_THRESHOLD value and SmallMap, not changing the current behaviour.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscr...@pdfbox.apache.org
For additional commands, e-mail: dev-h...@pdfbox.apache.org
