[
https://issues.apache.org/jira/browse/MYFACES-3477?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15190387#comment-15190387
]
Leonardo Uribe commented on MYFACES-3477:
-----------------------------------------
I have been checking this one and it could be an opportunity of improvement
here.
The use of ByteArrayOutputStream involves two arrays: one that is dynamically
expanded (initial size 32) and the other that is created when toByteArray()
method is called, copying the content from the internal created buffer. This is
inefficient because it uses the double or even more memory than it really needs.
The situations where ByteArrayOutputStream is used are:
- Serialize the component tree into a byte array
- Compress the serialized byte array.
- Encrypt the final array (a byte array is created).
The objective could be to create an algorithm that uses only 2 buffers and keep
track of the max size of the buffer, to avoid new allocations.
The performance gain could depend of the size of the view state. I would say
with the current PSS algorithm, in almost all cases the size is really small,
but that depends on the application, how the views are built and so on. It is
worth a shot.
> [perf] Review amount of byte [] arrays created during state saving
> -------------------------------------------------------------------
>
> Key: MYFACES-3477
> URL: https://issues.apache.org/jira/browse/MYFACES-3477
> Project: MyFaces Core
> Issue Type: Improvement
> Reporter: Martin Kočí
> Priority: Trivial
>
> instances of byte [] are created in:
> 1) org.apache.myfaces.shared.util.StateUtils.getAsByteArray(Object,
> ExternalContext)
> 2)
> org.apache.myfaces.shared.util.serial.DefaultSerialFactory.getObjectOutputStream
> 3) writer.writeObject(object); in
> org.apache.myfaces.shared.util.StateUtils.getAsByteArray
> 4) org.apache.myfaces.shared.util.StateUtils.encode(byte[])
> 5)
> org.apache.myfaces.shared.util.serial.DefaultSerialFactory.getObjectInputStream(InputStream)
> 6) org.apache.myfaces.shared.util.StateUtils.reconstruct(String,
> ExternalContext)
> those allocations are not performance problem; only in cumulative time are
> those methods #1 source of byte [] in myfaces - maybe can be some of them
> removed.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
