[Java] ArrowBuf bounds checking in getBytes/setBytes

[Benjamin Wilhelm <be...@knime.com>]

Hi all,

I have a very specific question about the `ArrowBuf` implementation in 
the Java project.
There is this constant `BOUNDS_CHECKING_ENABLED` which enables/disables 
checking of the bounds when accessing the buffer. However, the 
`getBytes` and `setBytes` implementations ignore this constant and 
always check the bounds. What is the reason for this? Wouldn't it make 
sense if each method respects this constant?

Thank you!
Benjamin Wilhelm

Re: [Java] ArrowBuf bounds checking in getBytes/setBytes

[Benjamin Wilhelm <be...@knime.com>]

Thank you Liya,

I will try to implement some benchmarks to see how much it impacts the 
performance for different numbers of bytes set each time. If I find 
enough evidence that bounds checking can hurt the performance I will 
open a PR.

Best,
Benjamin Wilhelm


On 15.10.20 14:07, Fan Liya wrote:
> Hi Benjamin,
> 
> Nice catch!
> 
> The code has been like this for quite some time.
> I think one reason is that the 'setBytes' and 'getBytes' APIs
> support manipulating data in large batches, so it is less
> performance-critical.
> 
> IMO, it is reasonable to respect the constant to improve performance.
> 
> Best,
> Liya Fan
> 
> 
> 
> On Thu, Oct 15, 2020 at 7:09 PM Benjamin Wilhelm <be...@knime.com>
> wrote:
> 
>> Hi all,
>>
>> I have a very specific question about the `ArrowBuf` implementation in
>> the Java project.
>> There is this constant `BOUNDS_CHECKING_ENABLED` which enables/disables
>> checking of the bounds when accessing the buffer. However, the
>> `getBytes` and `setBytes` implementations ignore this constant and
>> always check the bounds. What is the reason for this? Wouldn't it make
>> sense if each method respects this constant?
>>
>> Thank you!
>> Benjamin Wilhelm
>>
> 

Re: [Java] ArrowBuf bounds checking in getBytes/setBytes

[Fan Liya <li...@gmail.com>]

Hi Benjamin,

Nice catch!

The code has been like this for quite some time.
I think one reason is that the 'setBytes' and 'getBytes' APIs
support manipulating data in large batches, so it is less
performance-critical.

IMO, it is reasonable to respect the constant to improve performance.

Best,
Liya Fan



On Thu, Oct 15, 2020 at 7:09 PM Benjamin Wilhelm <be...@knime.com>
wrote:

> Hi all,
>
> I have a very specific question about the `ArrowBuf` implementation in
> the Java project.
> There is this constant `BOUNDS_CHECKING_ENABLED` which enables/disables
> checking of the bounds when accessing the buffer. However, the
> `getBytes` and `setBytes` implementations ignore this constant and
> always check the bounds. What is the reason for this? Wouldn't it make
> sense if each method respects this constant?
>
> Thank you!
> Benjamin Wilhelm
>