cell-serialisation

classic Classic list List threaded Threaded
4 messages Options
Reply | Threaded
Open this post in threaded view
|

cell-serialisation

Tobias Pietzsch
Hi Mark,

we just noticed the merge of the cell-serialization branch, that makes AbstractCell Serializable and therefore final field non-final.
Actually Stephan Saalfeld noticed, I would have missed it probably. It would be nice if such changes could be discussed before they are merged to master.
Did you do any tests to evaluate the performance impact of un-finalizing the fields.
What is this needed for and are there other alternatives to achieve it instead of making AbstractCell Serializable.

best regards,
Tobias
_______________________________________________
ImageJ-devel mailing list
[hidden email]
http://imagej.net/mailman/listinfo/imagej-devel
Reply | Threaded
Open this post in threaded view
|

Re: cell-serialisation

Mark Hiner
Hi Tobias,

The SCIFIOCell class needs to be serializable to allow the library we're using to persist opened cells to disk, and thus its superclass(es) needs to be serializable with non-final fields.

The only potential alternative I can think of would be to fully duplicate AbstractCell's fields and logic in SCIFIOCell... but that would still require the ArrayDataAccess to be serializable (do you see that as a potential problem, or only AbstractCell?).

In the merge commit I mentioned we ran the PerformanceBenchmark with/without implementing Serializable and saw no difference. We thought that was sufficient for merging the changes, but I'm sorry if we circumvented discussion.

If there are other tests you'd like to see run, or would like me to investigate duplicating AbstractCell's code, let me know.

Thanks,
Mark


On Wed, Jul 3, 2013 at 10:40 AM, Tobias Pietzsch <[hidden email]> wrote:
Hi Mark,

we just noticed the merge of the cell-serialization branch, that makes AbstractCell Serializable and therefore final field non-final.
Actually Stephan Saalfeld noticed, I would have missed it probably. It would be nice if such changes could be discussed before they are merged to master.
Did you do any tests to evaluate the performance impact of un-finalizing the fields.
What is this needed for and are there other alternatives to achieve it instead of making AbstractCell Serializable.

best regards,
Tobias


_______________________________________________
ImageJ-devel mailing list
[hidden email]
http://imagej.net/mailman/listinfo/imagej-devel
Reply | Threaded
Open this post in threaded view
|

Re: cell-serialisation

Tobias Pietzsch
Hi Mark

Ok, if it doesn't harm performance it's fine with me. I had missed the note in the merge commit...

best regards,
Tobias

On Jul 8, 2013, at 3:46 PM, Mark Hiner <[hidden email]> wrote:

Hi Tobias,

The SCIFIOCell class needs to be serializable to allow the library we're using to persist opened cells to disk, and thus its superclass(es) needs to be serializable with non-final fields.

The only potential alternative I can think of would be to fully duplicate AbstractCell's fields and logic in SCIFIOCell... but that would still require the ArrayDataAccess to be serializable (do you see that as a potential problem, or only AbstractCell?).

In the merge commit I mentioned we ran the PerformanceBenchmark with/without implementing Serializable and saw no difference. We thought that was sufficient for merging the changes, but I'm sorry if we circumvented discussion.

If there are other tests you'd like to see run, or would like me to investigate duplicating AbstractCell's code, let me know.

Thanks,
Mark


On Wed, Jul 3, 2013 at 10:40 AM, Tobias Pietzsch <[hidden email]> wrote:
Hi Mark,

we just noticed the merge of the cell-serialization branch, that makes AbstractCell Serializable and therefore final field non-final.
Actually Stephan Saalfeld noticed, I would have missed it probably. It would be nice if such changes could be discussed before they are merged to master.
Did you do any tests to evaluate the performance impact of un-finalizing the fields.
What is this needed for and are there other alternatives to achieve it instead of making AbstractCell Serializable.

best regards,
Tobias



_______________________________________________
ImageJ-devel mailing list
[hidden email]
http://imagej.net/mailman/listinfo/imagej-devel
Reply | Threaded
Open this post in threaded view
|

Re: cell-serialisation

Mark Hiner-2
Yeah I realized that it was in the merge commit but not the actual commit. Sorry for the confusion.


On Wed, Jul 10, 2013 at 10:42 AM, Tobias Pietzsch <[hidden email]> wrote:
Hi Mark

Ok, if it doesn't harm performance it's fine with me. I had missed the note in the merge commit...

best regards,
Tobias

On Jul 8, 2013, at 3:46 PM, Mark Hiner <[hidden email]> wrote:

Hi Tobias,

The SCIFIOCell class needs to be serializable to allow the library we're using to persist opened cells to disk, and thus its superclass(es) needs to be serializable with non-final fields.

The only potential alternative I can think of would be to fully duplicate AbstractCell's fields and logic in SCIFIOCell... but that would still require the ArrayDataAccess to be serializable (do you see that as a potential problem, or only AbstractCell?).

In the merge commit I mentioned we ran the PerformanceBenchmark with/without implementing Serializable and saw no difference. We thought that was sufficient for merging the changes, but I'm sorry if we circumvented discussion.

If there are other tests you'd like to see run, or would like me to investigate duplicating AbstractCell's code, let me know.

Thanks,
Mark


On Wed, Jul 3, 2013 at 10:40 AM, Tobias Pietzsch <[hidden email]> wrote:
Hi Mark,

we just noticed the merge of the cell-serialization branch, that makes AbstractCell Serializable and therefore final field non-final.
Actually Stephan Saalfeld noticed, I would have missed it probably. It would be nice if such changes could be discussed before they are merged to master.
Did you do any tests to evaluate the performance impact of un-finalizing the fields.
What is this needed for and are there other alternatives to achieve it instead of making AbstractCell Serializable.

best regards,
Tobias




_______________________________________________
ImageJ-devel mailing list
[hidden email]
http://imagej.net/mailman/listinfo/imagej-devel