On Monday, 30 March 2015 at 05:04:57 UTC, Rikki Cattermole wrote:
> On 30/03/2015 5:48 p.m., weaselcat wrote:
>> On Monday, 30 March 2015 at 04:35:44 UTC, Rikki Cattermole wrote:
>>> On 30/03/2015 5:25 p.m., Laeeth Isharc wrote:
>>>> On Monday, 30 March 2015 at 04:16:38 UTC, weaselcat wrote:
>>>>> On Monday, 30 March 2015 at 00:57:06 UTC, Walter Bright wrote:
>>>>>> http://blog.experimentalworks.net/2015/01/the-d-language-a-sweet-spot-between-python-and-c/
>>>>>>
>>>>>>
>>>>>>
>>>>>> Reddit:
>>>>>> https://www.reddit.com/r/programming/comments/30qqck/the_d_language_a_sweetspot_between_python_and_c/
>>>>>>
>>>>>>
>>>>>
>>>>> a lot of the people in the thread are unaware that D even has RAII
>>>>> like C++, and think it's just a GC language like java/etc. Maybe this
>>>>> is something worth mentioning more on the introduction?
>>>>
>>>> Also, there is a perception that you can't  use the standard library and
>>>> nicer language features if you do your own allocation and don't depend
>>>> on the GC.  A guy worrying about hygiene problems mixing GC and Raii
>>>> libraries.  Whereas most garbage is small and fine to use GC for in some
>>>> applications - only a subset of real time applications suffer from
>>>> generating gazillions of tiny objects.  It would be good to set out
>>>> somewhere what you lose as regards std library by insisting on using
>>>> nogc.  The point about std.algorithm should be made more prominent.
>>>
>>> I'm currently working on the forcing GC cleanup mechanism for my web
>>> server. I would like to add, that post GC disabled it can be forced to
>>> do a cleanup.
>>>
>>> But I would go a step further, do a force minimize of memory back to
>>> the OS and reserve e.g. 32mb. Really what would be nice is a,
>>> reserveMax function that and anything else is free'd back to the OS.
>>>
>>> The reserve, means that even if you are sloppy and end up using the GC
>>> in critical code, it won't matter. The memory is already allocated.
>>> Cleaning up can happen during non critical times. After all, if you
>>> are using more then e.g. 32mb in critical code, you are doing
>>> something wrong.
>>
>>
>> I actually use D for a pet project of mine(a game! ;) ) and this is what
>> I do. I leave the GC disabled and essentially just use it as a free
>> store(while not haphazardly abusing it,) and just manually clean it
>> during opportune times.
>>
>> It's also better to have a single pause for a large cleanup than many
>> small pauses, the overhead of actually scanning the memory will kill you.
>
> Atleast with web servers, a whole bunch of pauses can't be dealt with. But one large one, can easily be via load balances.

How about we (ie you, the language expert!) jot down a few more points  to later turn into a short but useful article on how to deal with the GC in practical  situations?

On 30/03/2015 6:43 p.m., Laeeth Isharc wrote:
> On Monday, 30 March 2015 at 05:04:57 UTC, Rikki Cattermole wrote:
>> On 30/03/2015 5:48 p.m., weaselcat wrote:
>>> On Monday, 30 March 2015 at 04:35:44 UTC, Rikki Cattermole wrote:
>>>> On 30/03/2015 5:25 p.m., Laeeth Isharc wrote:
>>>>> On Monday, 30 March 2015 at 04:16:38 UTC, weaselcat wrote:
>>>>>> On Monday, 30 March 2015 at 00:57:06 UTC, Walter Bright wrote:
>>>>>>> http://blog.experimentalworks.net/2015/01/the-d-language-a-sweet-spot-between-python-and-c/
>>>>>>>
>>>>>>>
>>>>>>>
>>>>>>>
>>>>>>> Reddit:
>>>>>>> https://www.reddit.com/r/programming/comments/30qqck/the_d_language_a_sweetspot_between_python_and_c/
>>>>>>>
>>>>>>>
>>>>>>>
>>>>>>
>>>>>> a lot of the people in the thread are unaware that D even has RAII
>>>>>> like C++, and think it's just a GC language like java/etc. Maybe this
>>>>>> is something worth mentioning more on the introduction?
>>>>>
>>>>> Also, there is a perception that you can't  use the standard
>>>>> library and
>>>>> nicer language features if you do your own allocation and don't depend
>>>>> on the GC.  A guy worrying about hygiene problems mixing GC and Raii
>>>>> libraries.  Whereas most garbage is small and fine to use GC for in
>>>>> some
>>>>> applications - only a subset of real time applications suffer from
>>>>> generating gazillions of tiny objects.  It would be good to set out
>>>>> somewhere what you lose as regards std library by insisting on using
>>>>> nogc.  The point about std.algorithm should be made more prominent.
>>>>
>>>> I'm currently working on the forcing GC cleanup mechanism for my web
>>>> server. I would like to add, that post GC disabled it can be forced to
>>>> do a cleanup.
>>>>
>>>> But I would go a step further, do a force minimize of memory back to
>>>> the OS and reserve e.g. 32mb. Really what would be nice is a,
>>>> reserveMax function that and anything else is free'd back to the OS.
>>>>
>>>> The reserve, means that even if you are sloppy and end up using the GC
>>>> in critical code, it won't matter. The memory is already allocated.
>>>> Cleaning up can happen during non critical times. After all, if you
>>>> are using more then e.g. 32mb in critical code, you are doing
>>>> something wrong.
>>>
>>>
>>> I actually use D for a pet project of mine(a game! ;) ) and this is what
>>> I do. I leave the GC disabled and essentially just use it as a free
>>> store(while not haphazardly abusing it,) and just manually clean it
>>> during opportune times.
>>>
>>> It's also better to have a single pause for a large cleanup than many
>>> small pauses, the overhead of actually scanning the memory will kill
>>> you.
>>
>> Atleast with web servers, a whole bunch of pauses can't be dealt with.
>> But one large one, can easily be via load balances.
>
> How about we (ie you, the language expert!) jot down a few more points
> to later turn into a short but useful article on how to deal with the GC
> in practical  situations?

I'm by far not a language expert, especially with manual memory management. I can only discuss what I've dealt with my own projects.

I have yet to get to the part where I have to actually try to be @nogc or pre allocate + buffers.

But there will be an article at some point. But only when it is ready to go public. I am sure there will be a lot of interest in an Apache equivalent web server in D. With shared library support.

>> How about we (ie you, the language expert!) jot down a few more points
>> to later turn into a short but useful article on how to deal with the GC
>> in practical  situations?
>
> I'm by far not a language expert, especially with manual memory management. I can only discuss what I've dealt with my own projects.
>
> I have yet to get to the part where I have to actually try to be @nogc or pre allocate + buffers.
>
> But there will be an article at some point. But only when it is ready to go public. I am sure there will be a lot of interest in an Apache equivalent web server in D. With shared library support.

I like my new oneplusone smart phone, but it doesn't lend itself to thoughtful expression.  Yes - understand that, but this bit is what I mean: " I can only discuss what I've dealt with my own projects".  One of the best things about Adam Ruppe's books and talks is the way he takes you on the journey of how he figured something out.  Humans learn as much by imitating those ahead of them (who seem human and not out of reach) as by book learning.  Understand wanting to wait before saying much, but at the moment the GC is one of those effective FUD factors even though my guess is it needn't be for many people who use it as an excuse not to look any further into D.

This is what there is currently:
http://wiki.dlang.org/Memory_Management
http://wiki.dlang.org/Instantiating_Class_Objects_Elsewhere_Than_the_GC_Heap