[
https://issues.apache.org/jira/browse/IO-355?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13488154#comment-13488154
]
Sebb edited comment on IO-355 at 10/31/12 7:55 PM:
---------------------------------------------------
See IO-203 where the skip() methods were added.
The problem is that some implementations of skip() - e.g. FileInputStream -
allow skipping past EOF.
That is why we use our own implementation using read() rather than delegating
to the implementation class.
Also, in the case of InputStream#skip(), the Javadoc says it throws:
bq. IOException - if the stream does not support seek or ...
An unconditional change to using skip from the implementation class could break
some applications.
was (Author: [email protected]):
In the case of InputStream#skip(), the Javadoc says it throws:
bq. IOException - if the stream does not support seek or ...
so that might explain why the code uses read() rather than seek().
The Javadoc for Reader#skip() does not mention whether this is a possibility.
[The two classes have different authors, which may explain why Reader#skip
throws IllegalArgumentException for negative arg but InputStream#skip() does
not]
An unconditional change to using skip from the implementation class could break
some applications, so this is not a trivial fix.
> IOUtils read() and skip() methods are performance hogs
> ------------------------------------------------------
>
> Key: IO-355
> URL: https://issues.apache.org/jira/browse/IO-355
> Project: Commons IO
> Issue Type: Bug
> Components: Utilities
> Affects Versions: 2.3, 2.4
> Reporter: Uli Bubenheimer
>
> IOUtils.skip(InputStream, long) and IOUtils.skip(Reader, long) have the worst
> possible performance as they always use read() on the input instead of using
> skip(). In many cases, using skip() from a subclass of InputStream is much
> faster than read(), as the skip() can be implemented via a disk seek.
> The IOUtils.skip() methods are used in the read() methods of IOUtils and
> their similarly named siblings, so they tend to bring down the performance of
> all reads that involve a skip.
> Case in point: I have observed this performance degradation with Java 7 on
> Windows 7. A series of consecutive reads on a large file on disk that
> involved skips changed my performance from 30 secs as my baseline to 10
> minutes after starting to use IOUtils.read().
--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators
For more information on JIRA, see: http://www.atlassian.com/software/jira
