quantlib-users — QuantLib users discussion list

Re: [Quantlib-users] How best to store historical yield curves?

[Luigi B. <lui...@gm...>]

Nope, objects exported through SWIG are not pickleable – if that's the word
:)

Luigi


On Tue, Nov 17, 2020 at 11:20 AM <ben...@ma...> wrote:

> From QuantlibXL you can serialize into XML. I have used this and it works
> quite well,  but that is probably not that useful if your client is not
> Excel.  From Python I don’t think that Quantlib supports serialization but
> I am not 100% sure on that. I did take a quick look at Pickle in python,
> but did not get to the testing stage.
>
>
>
> *From:* da...@el... <da...@el...>
> *Sent:* Tuesday, 17 November 2020 12:34 AM
> *To:* qua...@li...
> *Subject:* [Quantlib-users] How best to store historical yield curves?
>
>
>
> Hi everyone,
>
> My apologies if this has been covered before, but here goes:
>
> I’d like to store curves, to build up a history. Eg, store a history of
> Euribor 6m curves, and EONIA curves, constructed from historical data. I
> could simply serialize every object that each curve requires (ratehandlers
> etc) and rebuild as needed, but is there a more efficient way? A set of
> dates and discount factors perhaps? I don’t necessarily want to
> “re-bootstrap” every curve for each day.
>
>
>
> Thanks as always,
>
> David
>
>
>
> _______________________________________________
> QuantLib-users mailing list
> Qua...@li...
> https://lists.sourceforge.net/lists/listinfo/quantlib-users
>

Re: [Quantlib-users] How best to store historical yield curves?

[<ben...@ma...>]

>From QuantlibXL you can serialize into XML. I have used this and it works
quite well,  but that is probably not that useful if your client is not
Excel.  From Python I don’t think that Quantlib supports serialization but I
am not 100% sure on that. I did take a quick look at Pickle in python, but
did not get to the testing stage.

 

From: da...@el... <da...@el...> 
Sent: Tuesday, 17 November 2020 12:34 AM
To: qua...@li...
Subject: [Quantlib-users] How best to store historical yield curves?

 

Hi everyone,

My apologies if this has been covered before, but here goes:

I’d like to store curves, to build up a history. Eg, store a history of
Euribor 6m curves, and EONIA curves, constructed from historical data. I
could simply serialize every object that each curve requires (ratehandlers
etc) and rebuild as needed, but is there a more efficient way? A set of
dates and discount factors perhaps? I don’t necessarily want to
“re-bootstrap” every curve for each day.

 

Thanks as always,

David

Re: [Quantlib-users] DLL Load Failed

[Luigi B. <lui...@gm...>]

>From the file names and paths, it seems you have the 64-bit version of
QuantLib installed but spyder is using 32-bit python. May you check that?

Luigi


On Tue, Nov 17, 2020 at 3:52 AM Teddy O <tm...@ho...> wrote:

> I installed quantlib using "pip install QuantLib" and now I'm getting the
> below error. I'm using windows 10 and Spyder IDE. I don't see any dll file
> called _QuantLib, only _QuantLib.cp37-win_amd64.pyd.
>
>
> File "C:\Conda\lib\site-packages\QuantLib\QuantLib.py", line 13, in from .
> import _QuantLib
>
>
> ImportError: DLL load failed: The specified module could not be found.
>
>
> These are my path variables:
>
>
> C:\Conda\python38.zip
> C:\Conda\DLLs
> C:\Conda\lib
> C:\Conda
>
> C:\Conda\lib\site-packages
> C:\Conda\lib\site-packages\win32
> C:\Conda\lib\site-packages\win32\lib
> C:\Conda\lib\site-packages\Pythonwin
> C:\Conda\lib\site-packages\IPython\extensions
> C:\Users\name t\.ipython
>
> Has anyone experienced a similar problem?
>
> Thanks
>
>
> Sent from Outlook <http://aka.ms/weboutlook>
> _______________________________________________
> QuantLib-users mailing list
> Qua...@li...
> https://lists.sourceforge.net/lists/listinfo/quantlib-users
>

[Quantlib-users] DLL Load Failed

[Teddy O <tm...@ho...>]

I installed quantlib using "pip install QuantLib" and now I'm getting the below error. I'm using windows 10 and Spyder IDE. I don't see any dll file called _QuantLib, only _QuantLib.cp37-win_amd64.pyd.


File "C:\Conda\lib\site-packages\QuantLib\QuantLib.py", line 13, in from . import _QuantLib


ImportError: DLL load failed: The specified module could not be found.


These are my path variables:


C:\Conda\python38.zip
C:\Conda\DLLs
C:\Conda\lib
C:\Conda

C:\Conda\lib\site-packages
C:\Conda\lib\site-packages\win32
C:\Conda\lib\site-packages\win32\lib
C:\Conda\lib\site-packages\Pythonwin
C:\Conda\lib\site-packages\IPython\extensions
C:\Users\name t\.ipython

Has anyone experienced a similar problem?

Thanks



Sent from Outlook<http://aka.ms/weboutlook>

Re: [Quantlib-users] How best to store historical yield curves?

[<da...@el...>]

Thanks Luigi!

That seems to be exactly what I am after: to save the minimum data set that will let me efficiently re-create a curve in memory.

 

Best wishes,

 

David

 

From: Luigi Ballabio <lui...@gm...> 
Sent: Monday, 16 November 2020 13:49
To: da...@el...
Cc: QuantLib users <qua...@li...>
Subject: Re: [Quantlib-users] How best to store historical yield curves?

 

Hi David,

    calling the nodes() method on a bootstrapped curve will give you the set of dates and values resulting from the bootstrap. Depending on the type of your curve, they could be dates and discount factors, or dates and zero rates, or dates and forward rates.  You can build a curve with the same data and interpolation by feeding them back to the corresponding interpolated class; e.g., nodes from PiecewiseYieldCurve<Discount, LogCubic> would go into building an InterpolatedDiscountCurve<LogCubic>, and nodes from PiecewiseYieldCurve<ForwardRate, BackwardFlat> would go into an InterpolatedForwardCurve<BackwardFlat>.

 

Luigi

 

 

On Mon, Nov 16, 2020 at 2:36 PM <da...@el... <mailto:da...@el...> > wrote:

Hi everyone,

My apologies if this has been covered before, but here goes:

I’d like to store curves, to build up a history. Eg, store a history of Euribor 6m curves, and EONIA curves, constructed from historical data. I could simply serialize every object that each curve requires (ratehandlers etc) and rebuild as needed, but is there a more efficient way? A set of dates and discount factors perhaps? I don’t necessarily want to “re-bootstrap” every curve for each day.

 

Thanks as always,

David





_______________________________________________
QuantLib-users mailing list
Qua...@li... <mailto:Qua...@li...> 
https://lists.sourceforge.net/lists/listinfo/quantlib-users

Re: [Quantlib-users] How best to store historical yield curves?

[<da...@el...>]

Thanks Amine,

I had planned to run a one-time historical batch build that would run through one day at a time, collect that day’s data, bootstrap the curve and save it. Then maintain that history daily with new curves.

David

 

From: Amine Ifri <ami...@gm...> 
Sent: Monday, 16 November 2020 13:39
To: da...@el...
Cc: qua...@li...
Subject: Re: [Quantlib-users] How best to store historical yield curves?

 

Hi David,

 

If all you have is market instrument quotes then I would assume there is no way to avoid bootstrapping first, but if you already have discount factors for a set of dates, the DiscountCurve class has a constructor that takes dates and discount factors as inputs...

 

Hope this helps,

Amine Ifri





On 16 Nov 2020, at 13:35, da...@el... <mailto:da...@el...>  wrote:



Hi everyone,

My apologies if this has been covered before, but here goes:

I’d like to store curves, to build up a history. Eg, store a history of Euribor 6m curves, and EONIA curves, constructed from historical data. I could simply serialize every object that each curve requires (ratehandlers etc) and rebuild as needed, but is there a more efficient way? A set of dates and discount factors perhaps? I don’t necessarily want to “re-bootstrap” every curve for each day.

 

Thanks as always,

David







_______________________________________________
QuantLib-users mailing list
Qua...@li... <mailto:Qua...@li...> 
https://lists.sourceforge.net/lists/listinfo/quantlib-users

Re: [Quantlib-users] How best to store historical yield curves?

[Luigi B. <lui...@gm...>]

Hi David,
    calling the nodes() method on a bootstrapped curve will give you the
set of dates and values resulting from the bootstrap. Depending on the type
of your curve, they could be dates and discount factors, or dates and zero
rates, or dates and forward rates.  You can build a curve with the same
data and interpolation by feeding them back to the corresponding
interpolated class; e.g., nodes from PiecewiseYieldCurve<Discount,
LogCubic> would go into building an InterpolatedDiscountCurve<LogCubic>,
and nodes from PiecewiseYieldCurve<ForwardRate, BackwardFlat> would go into
an InterpolatedForwardCurve<BackwardFlat>.

Luigi


On Mon, Nov 16, 2020 at 2:36 PM <da...@el...> wrote:

> Hi everyone,
>
> My apologies if this has been covered before, but here goes:
>
> I’d like to store curves, to build up a history. Eg, store a history of
> Euribor 6m curves, and EONIA curves, constructed from historical data. I
> could simply serialize every object that each curve requires (ratehandlers
> etc) and rebuild as needed, but is there a more efficient way? A set of
> dates and discount factors perhaps? I don’t necessarily want to
> “re-bootstrap” every curve for each day.
>
>
>
> Thanks as always,
>
> David
>
>
>
>
> _______________________________________________
> QuantLib-users mailing list
> Qua...@li...
> https://lists.sourceforge.net/lists/listinfo/quantlib-users
>

Re: [Quantlib-users] How best to store historical yield curves?

[Amine I. <ami...@gm...>]

Hi David,

If all you have is market instrument quotes then I would assume there is no way to avoid bootstrapping first, but if you already have discount factors for a set of dates, the DiscountCurve class has a constructor that takes dates and discount factors as inputs...

Hope this helps,

Amine Ifri

> On 16 Nov 2020, at 13:35, da...@el... wrote:
> 
> 
> Hi everyone,
> 
> My apologies if this has been covered before, but here goes:
> 
> I’d like to store curves, to build up a history. Eg, store a history of Euribor 6m curves, and EONIA curves, constructed from historical data. I could simply serialize every object that each curve requires (ratehandlers etc) and rebuild as needed, but is there a more efficient way? A set of dates and discount factors perhaps? I don’t necessarily want to “re-bootstrap” every curve for each day.
>  
> Thanks as always,
> 
> David
> 
> 
> 
> _______________________________________________
> QuantLib-users mailing list
> Qua...@li...
> https://lists.sourceforge.net/lists/listinfo/quantlib-users

[Quantlib-users] How best to store historical yield curves?

[<da...@el...>]

Hi everyone,

My apologies if this has been covered before, but here goes:

I'd like to store curves, to build up a history. Eg, store a history of
Euribor 6m curves, and EONIA curves, constructed from historical data. I
could simply serialize every object that each curve requires (ratehandlers
etc) and rebuild as needed, but is there a more efficient way? A set of
dates and discount factors perhaps? I don't necessarily want to
"re-bootstrap" every curve for each day.

 

Thanks as always,

David

Re: [Quantlib-users] How to expose QuantLibAddin to Python?

[Eric E. <eri...@re...>]

I don't think it makes sense to load the XLL into Python.

One motivation for exporting the QuantLibAddin interface to a non-Excel 
platform such as C or C++ is that you can then exactly replicate the 
behavior of a spreadsheet in some other environment.  I have used 
exactly that for grid computing where the spreadsheet sent calculations 
off to the grid for processing.

But if you just want to call QuantLib from Python and you don't care 
about matching a spreadsheet then it would make more sense to use 
QuantLib's normal SWIG binding to Python.

You could export the QuantLibAddin interface to COM and I think we may 
have even done that at one point, I can't remember.  Not sure how you 
would go about calling pure QuantLib from COM.

Regards,
Eric

On 11/16/20 12:54 PM, da...@el... wrote:
>
> My question would be why would you want to use QuantLibXL in Python?
>
> As I understand it, the main point of the Excel interface is to create 
> a system which holds object state, as Excel (and some other 
> frameworks) don’t support this. Python is OO-ready, so doesn’t require 
> this ObjectHandler layer. All it would add is a lot of unnecessary 
> type conversion from Python -> C++ -> Excel datatypes.
>
> Since we are on the subject of generating interfaces, I was wondering 
> whether a COM interface would be possible: ie auto-generate the IDL 
> file from C++ class definitions to add the COM layer onto the existing 
> QL classes.
>
> Best wishes,
>
> David
>
> *From:*Eric Ehlers <eri...@re...>
> *Sent:* Monday, 16 November 2020 10:34
> *To:* SX L <han...@ho...>
> *Cc:* qua...@li...
> *Subject:* Re: [Quantlib-users] How to expose QuantLibAddin to Python?
>
> > I managed to load QuantLibXLL into Python with win32com.
>
> That's an interesting exercise to attempt.  An XLL is a DLL so I guess 
> there is nothing to prevent you from loading it as such into Python.  
> But the XLL's entry points are designed to be called from Excel so it 
> would be a hassle at best to use from Python.
>
> > However, what about Linux side? SWIG? Can gensrc generate C interfaces similar to 
> XL ones?
>
> The idea behind gensrc is that you could autogenerate source code for 
> the QuantLibAddin interface for any platform, Excel is just one 
> example.  There is a build for C++ but it only supports a subset of 
> the functionality that QuantLibXL supports.  So the C++ build is more 
> of a proof of concept.  It works on Windows and Linux and is 
> documented here:
>
> https://www.quantlib.org/quantlibaddin/build_cpp.html
>
> Once upon a time, we implemented a gensrc module to export the 
> QuantLibAddin interface to C, and that facilitated deployment to 
> platforms such as Python that talk C.  However that build has not been 
> maintained for a long time and it would not be a trivial task to 
> resurrect it.
>
> Regards,
> Eric
>
> On 11/15/20 5:13 PM, SX L wrote:
>
>     I managed to load QuantLibXLL into Python with win32com. However,
>     what about Linux side? SWIG? Can gensrc generate C interfaces
>     similar to XL ones?
>
>
>
>
>     _______________________________________________
>
>     QuantLib-users mailing list
>
>     Qua...@li...  <mailto:Qua...@li...>
>
>     https://lists.sourceforge.net/lists/listinfo/quantlib-users
>

Re: [Quantlib-users] How to expose QuantLibAddin to Python?

[<da...@el...>]

My question would be why would you want to use QuantLibXL in Python? 

As I understand it, the main point of the Excel interface is to create a
system which holds object state, as Excel (and some other frameworks) don't
support this. Python is OO-ready, so doesn't require this ObjectHandler
layer. All it would add is a lot of unnecessary type conversion from Python
-> C++ -> Excel datatypes.

 

Since we are on the subject of generating interfaces, I was wondering
whether a COM interface would be possible: ie auto-generate the IDL file
from C++ class definitions to add the COM layer onto the existing QL
classes. 

Best wishes,

 

David

 

From: Eric Ehlers <eri...@re...> 
Sent: Monday, 16 November 2020 10:34
To: SX L <han...@ho...>
Cc: qua...@li...
Subject: Re: [Quantlib-users] How to expose QuantLibAddin to Python?

 

> I managed to load QuantLibXLL into Python with win32com. 

 

That's an interesting exercise to attempt.  An XLL is a DLL so I guess there
is nothing to prevent you from loading it as such into Python.  But the
XLL's entry points are designed to be called from Excel so it would be a
hassle at best to use from Python.

 

> However, what about Linux side? SWIG? Can gensrc generate C interfaces
similar to XL ones?

 

The idea behind gensrc is that you could autogenerate source code for the
QuantLibAddin interface for any platform, Excel is just one example.  There
is a build for C++ but it only supports a subset of the functionality that
QuantLibXL supports.  So the C++ build is more of a proof of concept.  It
works on Windows and Linux and is documented here:

 

https://www.quantlib.org/quantlibaddin/build_cpp.html

 

Once upon a time, we implemented a gensrc module to export the QuantLibAddin
interface to C, and that facilitated deployment to platforms such as Python
that talk C.  However that build has not been maintained for a long time and
it would not be a trivial task to resurrect it.

 

Regards,
Eric

 

On 11/15/20 5:13 PM, SX L wrote:

I managed to load QuantLibXLL into Python with win32com. However, what about
Linux side? SWIG? Can gensrc generate C interfaces similar to XL ones?

 






_______________________________________________
QuantLib-users mailing list
Qua...@li...
<mailto:Qua...@li...> 
https://lists.sourceforge.net/lists/listinfo/quantlib-users

Re: [Quantlib-users] How to expose QuantLibAddin to Python?

[Eric E. <eri...@re...>]

 > I managed to load QuantLibXLL into Python with win32com.


That's an interesting exercise to attempt.  An XLL is a DLL so I guess 
there is nothing to prevent you from loading it as such into Python.  
But the XLL's entry points are designed to be called from Excel so it 
would be a hassle at best to use from Python.


> However, what about Linux side? SWIG? Can  gensrc generate C interfaces similar to XL ones?


The idea behind gensrc is that you could autogenerate source code for 
the QuantLibAddin interface for any platform, Excel is just one 
example.  There is a build for C++ but it only supports a subset of the 
functionality that QuantLibXL supports.  So the C++ build is more of a 
proof of concept.  It works on Windows and Linux and is documented here:


https://www.quantlib.org/quantlibaddin/build_cpp.html


Once upon a time, we implemented a gensrc module to export the 
QuantLibAddin interface to C, and that facilitated deployment to 
platforms such as Python that talk C.  However that build has not been 
maintained for a long time and it would not be a trivial task to 
resurrect it.


Regards,
Eric


On 11/15/20 5:13 PM, SX L wrote:
> I managed to load QuantLibXLL into Python with win32com. However, what 
> about Linux side? SWIG? Can gensrc generate C interfaces similar to XL 
> ones?
>
>
>
> _______________________________________________
> QuantLib-users mailing list
> Qua...@li...
> https://lists.sourceforge.net/lists/listinfo/quantlib-users

Re: [Quantlib-users] ObservableSettings is deleted early than Observables.

[Peter C. <pca...@gm...>]

That makes sense. But then, how does your proposed fix help in this
situation? To me, it looks like the ObservableSettings instance might
still get deleted before the observables?

On Sun, 15 Nov 2020 at 16:30, SX L <han...@ho...> wrote:
>
> My guess is that the shared_ptr of those non-static Observables are stored in another static variable such as OH::Repository. So it depends on which dtor between the two static variables will be called first.
>
> ________________________________
> From: Peter Caspers <pca...@gm...>
> Sent: Saturday, November 14, 2020 8:08 PM
> To: SX L <han...@ho...>
> Cc: Eric Ehlers <eri...@re...>; qua...@li... <qua...@li...>
> Subject: Re: [Quantlib-users] ObservableSettings is deleted early than Observables.
>
> Hi Philip, I am just curious and maybe the question is naive, but if
> the ObservableSettings instance is stored as a
> shared_ptr<ObservableSettings> within a static map, how can that
> instance be destroyed before any non-static Observable instance?
> Thanks, Peter
>
> On Sat, 14 Nov 2020 at 09:27, SX L <han...@ho...> wrote:
> >
> > I don't have a github account yet, and I am an old fashion coder still using svn. However, the change is simple as below. If I were to go into multithread later, I would dump or re-design Singleton pattern, which is notorious for multithread.
> >
> > Index: ql/patterns/singleton.hpp
> > ===================================================================
> > --- ql/patterns/singleton.hpp   (revision 130)
> > +++ ql/patterns/singleton.hpp   (working copy)
> > @@ -160,7 +160,7 @@
> >      T& Singleton<T>::instance() {
> >
> >          #if (QL_MANAGED == 0) && !defined(QL_SINGLETON_THREAD_SAFE_INIT)
> > -        static std::map<ThreadKey, ext::shared_ptr<T> > instances_;
> > +        static T instances_;
> >          #endif
> >
> >          // thread safe double checked locking pattern with atomic memory calls
> > @@ -185,9 +185,7 @@
> >          ThreadKey id = 0;
> >          #endif
> >
> > -        ext::shared_ptr<T>& instance = instances_[id];
> > -        if (!instance)
> > -            instance = ext::shared_ptr<T>(new T);
> > +        T* instance = &instances_;
> >
> >          #endif
> >
> > ________________________________
> > From: Eric Ehlers <eri...@re...>
> > Sent: Wednesday, November 11, 2020 4:35 AM
> > To: SX L <han...@ho...>
> > Cc: qua...@li... <qua...@li...>
> > Subject: Re: [Quantlib-users] ObservableSettings is deleted early than Observables.
> >
> >
> > Hi Philip,
> >
> > Would you be willing to submit a pull request?
> >
> > Regards,
> > Eric
> >
> > On 11/10/20 5:21 AM, SX L wrote:
> >
> > Hi Eric,
> >
> > Thanks for looking into this issue. Yes, I identified the issue in my windows x64 debug version. The issue could be hard to reproduce since my release version gives me normal exit code. It really depends on the sequence of calling dtors of those shared pointers at exit.
> >
> > So I would suggest moving away from shared pointers in Singletons. Create a singltonDestroyer or use static instances.
> >
> > Get Outlook for Android
> >
> > ________________________________
> > From: Eric Ehlers <eri...@re...>
> > Sent: Monday, November 9, 2020, 11:42 PM
> > To: SX L
> > Cc: qua...@li...
> > Subject: Re: [Quantlib-users] ObservableSettings is deleted early than Observables.
> >
> > Hi Philip,
> >
> > I tried unsuccessfully to recreate the error.  You did not say which operating system you are using but you mention "debug version" which sounds like visual studio so I tested it on Windows - I'm using Visual Studio 2019 on 64-bit WIndows 10.  I built it using "Debug (static runtime)" x64 and I ran ClientCppDemo and it behaves as I expected (no access violation).  Anything else I can try to recreate the error?
> >
> > I have not examined the code but your remarks sound sensible to me.  What resolution would you suggest?
> >
> > Regards,
> > Eric
> >
> > On 10/7/20 6:43 AM, SX L wrote:
> >
> > I was trying the Quantlib Addin. The QLADemo exits abnormally in the debug version with an access violation exception. It turned out to be the case that ObervableSettings is deleted earlier than some observer and observables at exit. So in the dtor of the observer, ObservableSettings is called to check if update is deferred, hence causing the exception. I feel it's unsafe to use shared pointers inside a singleton. Any suggestion? Thanks.
> >
> > Regards,
> >
> > Philip
> >
> >
> > Get Outlook for Android
> >
> >
> > _______________________________________________
> > QuantLib-users mailing list
> > Qua...@li...
> > https://lists.sourceforge.net/lists/listinfo/quantlib-users
> >
> >
> > _______________________________________________
> > QuantLib-users mailing list
> > Qua...@li...
> > https://lists.sourceforge.net/lists/listinfo/quantlib-users

[Quantlib-users] How to expose QuantLibAddin to Python?

[SX L <han...@ho...>]

I managed to load QuantLibXLL into Python with win32com. However, what about Linux side? SWIG? Can gensrc generate C interfaces similar to XL ones?

Re: [Quantlib-users] ObservableSettings is deleted early than Observables.

[SX L <han...@ho...>]

My guess is that the shared_ptr of those non-static Observables are stored in another static variable such as OH::Repository. So it depends on which dtor between the two static variables will be called first.

________________________________
From: Peter Caspers <pca...@gm...>
Sent: Saturday, November 14, 2020 8:08 PM
To: SX L <han...@ho...>
Cc: Eric Ehlers <eri...@re...>; qua...@li... <qua...@li...>
Subject: Re: [Quantlib-users] ObservableSettings is deleted early than Observables.

Hi Philip, I am just curious and maybe the question is naive, but if
the ObservableSettings instance is stored as a
shared_ptr<ObservableSettings> within a static map, how can that
instance be destroyed before any non-static Observable instance?
Thanks, Peter

On Sat, 14 Nov 2020 at 09:27, SX L <han...@ho...> wrote:
>
> I don't have a github account yet, and I am an old fashion coder still using svn. However, the change is simple as below. If I were to go into multithread later, I would dump or re-design Singleton pattern, which is notorious for multithread.
>
> Index: ql/patterns/singleton.hpp
> ===================================================================
> --- ql/patterns/singleton.hpp   (revision 130)
> +++ ql/patterns/singleton.hpp   (working copy)
> @@ -160,7 +160,7 @@
>      T& Singleton<T>::instance() {
>
>          #if (QL_MANAGED == 0) && !defined(QL_SINGLETON_THREAD_SAFE_INIT)
> -        static std::map<ThreadKey, ext::shared_ptr<T> > instances_;
> +        static T instances_;
>          #endif
>
>          // thread safe double checked locking pattern with atomic memory calls
> @@ -185,9 +185,7 @@
>          ThreadKey id = 0;
>          #endif
>
> -        ext::shared_ptr<T>& instance = instances_[id];
> -        if (!instance)
> -            instance = ext::shared_ptr<T>(new T);
> +        T* instance = &instances_;
>
>          #endif
>
> ________________________________
> From: Eric Ehlers <eri...@re...>
> Sent: Wednesday, November 11, 2020 4:35 AM
> To: SX L <han...@ho...>
> Cc: qua...@li... <qua...@li...>
> Subject: Re: [Quantlib-users] ObservableSettings is deleted early than Observables.
>
>
> Hi Philip,
>
> Would you be willing to submit a pull request?
>
> Regards,
> Eric
>
> On 11/10/20 5:21 AM, SX L wrote:
>
> Hi Eric,
>
> Thanks for looking into this issue. Yes, I identified the issue in my windows x64 debug version. The issue could be hard to reproduce since my release version gives me normal exit code. It really depends on the sequence of calling dtors of those shared pointers at exit.
>
> So I would suggest moving away from shared pointers in Singletons. Create a singltonDestroyer or use static instances.
>
> Get Outlook for Android
>
> ________________________________
> From: Eric Ehlers <eri...@re...>
> Sent: Monday, November 9, 2020, 11:42 PM
> To: SX L
> Cc: qua...@li...
> Subject: Re: [Quantlib-users] ObservableSettings is deleted early than Observables.
>
> Hi Philip,
>
> I tried unsuccessfully to recreate the error.  You did not say which operating system you are using but you mention "debug version" which sounds like visual studio so I tested it on Windows - I'm using Visual Studio 2019 on 64-bit WIndows 10.  I built it using "Debug (static runtime)" x64 and I ran ClientCppDemo and it behaves as I expected (no access violation).  Anything else I can try to recreate the error?
>
> I have not examined the code but your remarks sound sensible to me.  What resolution would you suggest?
>
> Regards,
> Eric
>
> On 10/7/20 6:43 AM, SX L wrote:
>
> I was trying the Quantlib Addin. The QLADemo exits abnormally in the debug version with an access violation exception. It turned out to be the case that ObervableSettings is deleted earlier than some observer and observables at exit. So in the dtor of the observer, ObservableSettings is called to check if update is deferred, hence causing the exception. I feel it's unsafe to use shared pointers inside a singleton. Any suggestion? Thanks.
>
> Regards,
>
> Philip
>
>
> Get Outlook for Android
>
>
> _______________________________________________
> QuantLib-users mailing list
> Qua...@li...
> https://lists.sourceforge.net/lists/listinfo/quantlib-users
>
>
> _______________________________________________
> QuantLib-users mailing list
> Qua...@li...
> https://lists.sourceforge.net/lists/listinfo/quantlib-users

Re: [Quantlib-users] ObservableSettings is deleted early than Observables.

[Peter C. <pca...@gm...>]

Hi Philip, I am just curious and maybe the question is naive, but if
the ObservableSettings instance is stored as a
shared_ptr<ObservableSettings> within a static map, how can that
instance be destroyed before any non-static Observable instance?
Thanks, Peter

On Sat, 14 Nov 2020 at 09:27, SX L <han...@ho...> wrote:
>
> I don't have a github account yet, and I am an old fashion coder still using svn. However, the change is simple as below. If I were to go into multithread later, I would dump or re-design Singleton pattern, which is notorious for multithread.
>
> Index: ql/patterns/singleton.hpp
> ===================================================================
> --- ql/patterns/singleton.hpp   (revision 130)
> +++ ql/patterns/singleton.hpp   (working copy)
> @@ -160,7 +160,7 @@
>      T& Singleton<T>::instance() {
>
>          #if (QL_MANAGED == 0) && !defined(QL_SINGLETON_THREAD_SAFE_INIT)
> -        static std::map<ThreadKey, ext::shared_ptr<T> > instances_;
> +        static T instances_;
>          #endif
>
>          // thread safe double checked locking pattern with atomic memory calls
> @@ -185,9 +185,7 @@
>          ThreadKey id = 0;
>          #endif
>
> -        ext::shared_ptr<T>& instance = instances_[id];
> -        if (!instance)
> -            instance = ext::shared_ptr<T>(new T);
> +        T* instance = &instances_;
>
>          #endif
>
> ________________________________
> From: Eric Ehlers <eri...@re...>
> Sent: Wednesday, November 11, 2020 4:35 AM
> To: SX L <han...@ho...>
> Cc: qua...@li... <qua...@li...>
> Subject: Re: [Quantlib-users] ObservableSettings is deleted early than Observables.
>
>
> Hi Philip,
>
> Would you be willing to submit a pull request?
>
> Regards,
> Eric
>
> On 11/10/20 5:21 AM, SX L wrote:
>
> Hi Eric,
>
> Thanks for looking into this issue. Yes, I identified the issue in my windows x64 debug version. The issue could be hard to reproduce since my release version gives me normal exit code. It really depends on the sequence of calling dtors of those shared pointers at exit.
>
> So I would suggest moving away from shared pointers in Singletons. Create a singltonDestroyer or use static instances.
>
> Get Outlook for Android
>
> ________________________________
> From: Eric Ehlers <eri...@re...>
> Sent: Monday, November 9, 2020, 11:42 PM
> To: SX L
> Cc: qua...@li...
> Subject: Re: [Quantlib-users] ObservableSettings is deleted early than Observables.
>
> Hi Philip,
>
> I tried unsuccessfully to recreate the error.  You did not say which operating system you are using but you mention "debug version" which sounds like visual studio so I tested it on Windows - I'm using Visual Studio 2019 on 64-bit WIndows 10.  I built it using "Debug (static runtime)" x64 and I ran ClientCppDemo and it behaves as I expected (no access violation).  Anything else I can try to recreate the error?
>
> I have not examined the code but your remarks sound sensible to me.  What resolution would you suggest?
>
> Regards,
> Eric
>
> On 10/7/20 6:43 AM, SX L wrote:
>
> I was trying the Quantlib Addin. The QLADemo exits abnormally in the debug version with an access violation exception. It turned out to be the case that ObervableSettings is deleted earlier than some observer and observables at exit. So in the dtor of the observer, ObservableSettings is called to check if update is deferred, hence causing the exception. I feel it's unsafe to use shared pointers inside a singleton. Any suggestion? Thanks.
>
> Regards,
>
> Philip
>
>
> Get Outlook for Android
>
>
> _______________________________________________
> QuantLib-users mailing list
> Qua...@li...
> https://lists.sourceforge.net/lists/listinfo/quantlib-users
>
>
> _______________________________________________
> QuantLib-users mailing list
> Qua...@li...
> https://lists.sourceforge.net/lists/listinfo/quantlib-users

Re: [Quantlib-users] ObservableSettings is deleted early than Observables.

[Eric E. <eri...@re...>]

OK, I will take a crack at it before the next release.  I hope that I 
can recreate the error somehow, otherwise it will be hard to test the fix.


Kind Regards,
Eric


On 11/14/20 9:25 AM, SX L wrote:
> I don't have a github account yet, and I am an old fashion coder still 
> using svn. However, the change is simple as below. If I were to go 
> into multithread later, I would dump or re-design Singleton pattern, 
> which is notorious for multithread.
>
> Index: ql/patterns/singleton.hpp
> ===================================================================
> --- ql/patterns/singleton.hpp   (revision 130)
> +++ ql/patterns/singleton.hpp   (working copy)
> @@ -160,7 +160,7 @@
>      T& Singleton<T>::instance() {
>
>          #if (QL_MANAGED == 0) && !defined(QL_SINGLETON_THREAD_SAFE_INIT)
> -        static std::map<ThreadKey, ext::shared_ptr<T> > instances_;
> +        static T instances_;
>          #endif
>
>          // thread safe double checked locking pattern with atomic 
> memory calls
> @@ -185,9 +185,7 @@
>          ThreadKey id = 0;
>          #endif
>
> -        ext::shared_ptr<T>& instance = instances_[id];
> -        if (!instance)
> -            instance = ext::shared_ptr<T>(new T);
> +        T* instance = &instances_;
>
>          #endif
>
> ------------------------------------------------------------------------
> *From:* Eric Ehlers <eri...@re...>
> *Sent:* Wednesday, November 11, 2020 4:35 AM
> *To:* SX L <han...@ho...>
> *Cc:* qua...@li... 
> <qua...@li...>
> *Subject:* Re: [Quantlib-users] ObservableSettings is deleted early 
> than Observables.
>
> Hi Philip,
>
> Would you be willing to submit a pull request?
>
> Regards,
> Eric
>
> On 11/10/20 5:21 AM, SX L wrote:
>> Hi Eric,
>>
>> Thanks for looking into this issue. Yes, I identified the issue in my 
>> windows x64 debug version. The issue could be hard to reproduce since 
>> my release version gives me normal exit code. It really depends on 
>> the sequence of calling dtors of those shared pointers at exit.
>>
>> So I would suggest moving away from shared pointers in Singletons. 
>> Create a singltonDestroyer or use static instances.
>>
>> Get Outlook for Android <https://aka.ms/ghei36>
>>
>> ------------------------------------------------------------------------
>> *From:* Eric Ehlers <eri...@re...> 
>> <mailto:eri...@re...>
>> *Sent:* Monday, November 9, 2020, 11:42 PM
>> *To:* SX L
>> *Cc:* qua...@li... 
>> <mailto:qua...@li...>
>> *Subject:* Re: [Quantlib-users] ObservableSettings is deleted early 
>> than Observables.
>>
>> Hi Philip,
>>
>> I tried unsuccessfully to recreate the error.  You did not say which 
>> operating system you are using but you mention "debug version" which 
>> sounds like visual studio so I tested it on Windows - I'm using 
>> Visual Studio 2019 on 64-bit WIndows 10.  I built it using "Debug 
>> (static runtime)" x64 and I ran ClientCppDemo and it behaves as I 
>> expected (no access violation).  Anything else I can try to recreate 
>> the error?
>>
>> I have not examined the code but your remarks sound sensible to me.  
>> What resolution would you suggest?
>>
>> Regards,
>> Eric
>>
>> On 10/7/20 6:43 AM, SX L wrote:
>>> I was trying the Quantlib Addin. The QLADemo exits abnormally in the 
>>> debug version with an access violation exception. It turned out to 
>>> be the case that ObervableSettings is deleted earlier than some 
>>> observer and observables at exit. So in the dtor of the observer, 
>>> ObservableSettings is called to check if update is deferred, hence 
>>> causing the exception. I feel it's unsafe to use shared pointers 
>>> inside a singleton. Any suggestion? Thanks.
>>>
>>> Regards,
>>>
>>> Philip
>>>
>>>
>>> Get Outlook for Android <https://aka.ms/ghei36>
>>>
>>>
>>> _______________________________________________
>>> QuantLib-users mailing list
>>> Qua...@li...  <mailto:Qua...@li...>
>>> https://lists.sourceforge.net/lists/listinfo/quantlib-users
>>

Re: [Quantlib-users] ObservableSettings is deleted early than Observables.

[SX L <han...@ho...>]

I don't have a github account yet, and I am an old fashion coder still using svn. However, the change is simple as below. If I were to go into multithread later, I would dump or re-design Singleton pattern, which is notorious for multithread.

Index: ql/patterns/singleton.hpp
===================================================================
--- ql/patterns/singleton.hpp   (revision 130)
+++ ql/patterns/singleton.hpp   (working copy)
@@ -160,7 +160,7 @@
     T& Singleton<T>::instance() {

         #if (QL_MANAGED == 0) && !defined(QL_SINGLETON_THREAD_SAFE_INIT)
-        static std::map<ThreadKey, ext::shared_ptr<T> > instances_;
+        static T instances_;
         #endif

         // thread safe double checked locking pattern with atomic memory calls
@@ -185,9 +185,7 @@
         ThreadKey id = 0;
         #endif

-        ext::shared_ptr<T>& instance = instances_[id];
-        if (!instance)
-            instance = ext::shared_ptr<T>(new T);
+        T* instance = &instances_;

         #endif

________________________________
From: Eric Ehlers <eri...@re...>
Sent: Wednesday, November 11, 2020 4:35 AM
To: SX L <han...@ho...>
Cc: qua...@li... <qua...@li...>
Subject: Re: [Quantlib-users] ObservableSettings is deleted early than Observables.


Hi Philip,

Would you be willing to submit a pull request?

Regards,
Eric

On 11/10/20 5:21 AM, SX L wrote:
Hi Eric,

Thanks for looking into this issue. Yes, I identified the issue in my windows x64 debug version. The issue could be hard to reproduce since my release version gives me normal exit code. It really depends on the sequence of calling dtors of those shared pointers at exit.

So I would suggest moving away from shared pointers in Singletons. Create a singltonDestroyer or use static instances.

Get Outlook for Android<https://aka.ms/ghei36>

________________________________
From: Eric Ehlers <eri...@re...><mailto:eri...@re...>
Sent: Monday, November 9, 2020, 11:42 PM
To: SX L
Cc: qua...@li...<mailto:qua...@li...>
Subject: Re: [Quantlib-users] ObservableSettings is deleted early than Observables.


Hi Philip,

I tried unsuccessfully to recreate the error.  You did not say which operating system you are using but you mention "debug version" which sounds like visual studio so I tested it on Windows - I'm using Visual Studio 2019 on 64-bit WIndows 10.  I built it using "Debug (static runtime)" x64 and I ran ClientCppDemo and it behaves as I expected (no access violation).  Anything else I can try to recreate the error?

I have not examined the code but your remarks sound sensible to me.  What resolution would you suggest?

Regards,
Eric

On 10/7/20 6:43 AM, SX L wrote:
I was trying the Quantlib Addin. The QLADemo exits abnormally in the debug version with an access violation exception. It turned out to be the case that ObervableSettings is deleted earlier than some observer and observables at exit. So in the dtor of the observer, ObservableSettings is called to check if update is deferred, hence causing the exception. I feel it's unsafe to use shared pointers inside a singleton. Any suggestion? Thanks.

Regards,

Philip


Get Outlook for Android<https://aka.ms/ghei36>




_______________________________________________
QuantLib-users mailing list
Qua...@li...<mailto:Qua...@li...>
https://lists.sourceforge.net/lists/listinfo/quantlib-users

Re: [Quantlib-users] RiskyBond in Python?

[Luigi B. <lui...@gm...>]

Hello Jian,
    you can open a new issue at
https://github.com/lballabio/QuantLib-SWIG/issues asking for the class to
be exported.  It's likely that someone will pick it up for next release.

Luigi


On Thu, Nov 12, 2020 at 2:44 AM jian Xu <jia...@gm...> wrote:

> Thank you so much for the detailed explanation, Jack.  I don't have
> the development environment set up.  So let me think about which way
> to go.  But generally, I do think this is a useful class to have in
> Python.
>
> On Wed, Nov 11, 2020 at 6:27 PM Jack G <jac...@gm...> wrote:
> >
> > Hi Jian,
> >
> > QuantLib relies on another project, which auto-generates interface code,
> to export functionality to python and other languages -
> https://github.com/lballabio/QuantLib-SWIG. The reason you can't see
> fixed/floating risky bond in the python wrap is because an interface for it
> hasn't been included in this file:
> https://github.com/lballabio/QuantLib-SWIG/blob/master/SWIG/bonds.i
> >
> > If you want to have a go at adding it yourself (it's an instructive
> process), you will need to:
> > - fork the QuantLib-SWIG repo, and locally add the code to the file I
> mentioned above, following the other examples in the file
> > - test your changes, which requires a few sub-steps (the first time,
> this process will take several hours to set up and run):
> >   a) build QuantLib v1.20 locally from source as described here:
> https://www.quantlib.org/install.shtml (if you're using Windows you can
> try Visual Studio, but personally I find it easiest to do it using Windows
> Docker Desktop inside a Docker container in a linux environment)
> >   b) build QuantLib-SWIG v1.20 with your modifications
> >   c) test the code changes locally using Python
> > - when the wrapper is working, make a pull request with your changes
> into the QuantLib-SWIG repo, and Luigi will merge your changes after the
> automated tests pass
> > - optionally, you might consider adding documentation for the new python
> classes to the repo that generates the online documentation shown here:
> https://quantlib-python-docs.readthedocs.io/en/latest/
> >
> > If this sounds too ambitious for you, let me know and I can make the
> changes to expose the class for you, as it looks like a useful class to
> have available.
> >
> > Best,
> > Jack
> >
> > On Thu, Nov 12, 2020 at 5:31 AM jian Xu <jia...@gm...> wrote:
> >>
> >> Hi,
> >>
> >> There is a RiskyBond class under experimental/credit in the C++ repo.
> >> However, when I'm using the Python interface, it seems that it's not
> >> there.  My question is, how to get the RiskyBond in Python?  Or is
> >> there a reason that it's not exposed?
> >>
> >> More generally, I am trying to calibrate a survival probability curve
> >> of a fixed rate bond.  And I think RiskyBond is the correct one to
> >> use.  If it's not in Python, what's the proper object for this task?
> >> Thanks.
> >>
> >> Jian
> >>
> >>
> >> _______________________________________________
> >> QuantLib-users mailing list
> >> Qua...@li...
> >> https://lists.sourceforge.net/lists/listinfo/quantlib-users
>
>
> _______________________________________________
> QuantLib-users mailing list
> Qua...@li...
> https://lists.sourceforge.net/lists/listinfo/quantlib-users
>

Re: [Quantlib-users] RiskyBond in Python?

[jian Xu <jia...@gm...>]

Thank you so much for the detailed explanation, Jack.  I don't have
the development environment set up.  So let me think about which way
to go.  But generally, I do think this is a useful class to have in
Python.

On Wed, Nov 11, 2020 at 6:27 PM Jack G <jac...@gm...> wrote:
>
> Hi Jian,
>
> QuantLib relies on another project, which auto-generates interface code, to export functionality to python and other languages - https://github.com/lballabio/QuantLib-SWIG. The reason you can't see fixed/floating risky bond in the python wrap is because an interface for it hasn't been included in this file: https://github.com/lballabio/QuantLib-SWIG/blob/master/SWIG/bonds.i
>
> If you want to have a go at adding it yourself (it's an instructive process), you will need to:
> - fork the QuantLib-SWIG repo, and locally add the code to the file I mentioned above, following the other examples in the file
> - test your changes, which requires a few sub-steps (the first time, this process will take several hours to set up and run):
>   a) build QuantLib v1.20 locally from source as described here: https://www.quantlib.org/install.shtml (if you're using Windows you can try Visual Studio, but personally I find it easiest to do it using Windows Docker Desktop inside a Docker container in a linux environment)
>   b) build QuantLib-SWIG v1.20 with your modifications
>   c) test the code changes locally using Python
> - when the wrapper is working, make a pull request with your changes into the QuantLib-SWIG repo, and Luigi will merge your changes after the automated tests pass
> - optionally, you might consider adding documentation for the new python classes to the repo that generates the online documentation shown here: https://quantlib-python-docs.readthedocs.io/en/latest/
>
> If this sounds too ambitious for you, let me know and I can make the changes to expose the class for you, as it looks like a useful class to have available.
>
> Best,
> Jack
>
> On Thu, Nov 12, 2020 at 5:31 AM jian Xu <jia...@gm...> wrote:
>>
>> Hi,
>>
>> There is a RiskyBond class under experimental/credit in the C++ repo.
>> However, when I'm using the Python interface, it seems that it's not
>> there.  My question is, how to get the RiskyBond in Python?  Or is
>> there a reason that it's not exposed?
>>
>> More generally, I am trying to calibrate a survival probability curve
>> of a fixed rate bond.  And I think RiskyBond is the correct one to
>> use.  If it's not in Python, what's the proper object for this task?
>> Thanks.
>>
>> Jian
>>
>>
>> _______________________________________________
>> QuantLib-users mailing list
>> Qua...@li...
>> https://lists.sourceforge.net/lists/listinfo/quantlib-users

Re: [Quantlib-users] RiskyBond in Python?

[Jack G <jac...@gm...>]

Hi Jian,

QuantLib relies on another project, which auto-generates interface code, to
export functionality to python and other languages -
https://github.com/lballabio/QuantLib-SWIG. The reason you can't see
fixed/floating risky bond in the python wrap is because an interface for it
hasn't been included in this file:
https://github.com/lballabio/QuantLib-SWIG/blob/master/SWIG/bonds.i

If you want to have a go at adding it yourself (it's an instructive
process), you will need to:
- fork the QuantLib-SWIG repo, and locally add the code to the file I
mentioned above, following the other examples in the file
- test your changes, which requires a few sub-steps (the first time, this
process will take several hours to set up and run):
  a) build QuantLib v1.20 locally from source as described here:
https://www.quantlib.org/install.shtml (if you're using Windows you can try
Visual Studio, but personally I find it easiest to do it using Windows
Docker Desktop inside a Docker container in a linux environment)
  b) build QuantLib-SWIG v1.20 with your modifications
  c) test the code changes locally using Python
- when the wrapper is working, make a pull request with your changes into
the QuantLib-SWIG repo, and Luigi will merge your changes after the
automated tests pass
- optionally, you might consider adding documentation for the new python
classes to the repo that generates the online documentation shown here:
https://quantlib-python-docs.readthedocs.io/en/latest/

If this sounds too ambitious for you, let me know and I can make the
changes to expose the class for you, as it looks like a useful class to
have available.

Best,
Jack

On Thu, Nov 12, 2020 at 5:31 AM jian Xu <jia...@gm...> wrote:

> Hi,
>
> There is a RiskyBond class under experimental/credit in the C++ repo.
> However, when I'm using the Python interface, it seems that it's not
> there.  My question is, how to get the RiskyBond in Python?  Or is
> there a reason that it's not exposed?
>
> More generally, I am trying to calibrate a survival probability curve
> of a fixed rate bond.  And I think RiskyBond is the correct one to
> use.  If it's not in Python, what's the proper object for this task?
> Thanks.
>
> Jian
>
>
> _______________________________________________
> QuantLib-users mailing list
> Qua...@li...
> https://lists.sourceforge.net/lists/listinfo/quantlib-users
>

[Quantlib-users] RiskyBond in Python?

[jian Xu <jia...@gm...>]

Hi,

There is a RiskyBond class under experimental/credit in the C++ repo.
However, when I'm using the Python interface, it seems that it's not
there.  My question is, how to get the RiskyBond in Python?  Or is
there a reason that it's not exposed?

More generally, I am trying to calibrate a survival probability curve
of a fixed rate bond.  And I think RiskyBond is the correct one to
use.  If it's not in Python, what's the proper object for this task?
Thanks.

Jian

Re: [Quantlib-users] ObservableSettings is deleted early than Observables.

[Eric E. <eri...@re...>]

Hi Philip,

Would you be willing to submit a pull request?

Regards,
Eric

On 11/10/20 5:21 AM, SX L wrote:
> Hi Eric,
>
> Thanks for looking into this issue. Yes, I identified the issue in my 
> windows x64 debug version. The issue could be hard to reproduce since 
> my release version gives me normal exit code. It really depends on the 
> sequence of calling dtors of those shared pointers at exit.
>
> So I would suggest moving away from shared pointers in Singletons. 
> Create a singltonDestroyer or use static instances.
>
> Get Outlook for Android <https://aka.ms/ghei36>
>
> ------------------------------------------------------------------------
> *From:* Eric Ehlers <eri...@re...>
> *Sent:* Monday, November 9, 2020, 11:42 PM
> *To:* SX L
> *Cc:* qua...@li...
> *Subject:* Re: [Quantlib-users] ObservableSettings is deleted early 
> than Observables.
>
> Hi Philip,
>
> I tried unsuccessfully to recreate the error.  You did not say which 
> operating system you are using but you mention "debug version" which 
> sounds like visual studio so I tested it on Windows - I'm using Visual 
> Studio 2019 on 64-bit WIndows 10.  I built it using "Debug (static 
> runtime)" x64 and I ran ClientCppDemo and it behaves as I expected (no 
> access violation).  Anything else I can try to recreate the error?
>
> I have not examined the code but your remarks sound sensible to me.  
> What resolution would you suggest?
>
> Regards,
> Eric
>
> On 10/7/20 6:43 AM, SX L wrote:
>> I was trying the Quantlib Addin. The QLADemo exits abnormally in the 
>> debug version with an access violation exception. It turned out to be 
>> the case that ObervableSettings is deleted earlier than some observer 
>> and observables at exit. So in the dtor of the observer, 
>> ObservableSettings is called to check if update is deferred, hence 
>> causing the exception. I feel it's unsafe to use shared pointers 
>> inside a singleton. Any suggestion? Thanks.
>>
>> Regards,
>>
>> Philip
>>
>>
>> Get Outlook for Android <https://aka.ms/ghei36>
>>
>>
>> _______________________________________________
>> QuantLib-users mailing list
>> Qua...@li...
>> https://lists.sourceforge.net/lists/listinfo/quantlib-users
>

[Quantlib-users] QuantLibXL / ObjectHandler 1.20 Released

[Eric E. <eri...@re...>]

QuantLibXL / ObjectHandler 1.20 Released

QuantLibXL, QuantLibAddin, ObjectHandler, and gensrc version
1.20 have been released and are available for download:
https://bintray.com/quantlib/releases/QuantLibXL/1.20#files

QuantLibAddin
http://www.quantlibaddin.org

QuantLibAddin exports the QuantLib interface to a variety
of end user platforms.

QuantLibXL
http://www.quantlibxl.org

QuantLibXL is the implementation of QuantLibAddin for
Microsoft Excel.  The QuantLibXL project includes a binary
release comprising a compiled Addin and example
workbooks.

ObjectHandler
http://www.objecthandler.org

ObjectHandler implements a repository where objects can be
stored, shared, updated, interrogated, and destroyed.  This
facilitates object orientation in procedural environments
such as spreadsheets.

Re: [Quantlib-users] ObservableSettings is deleted early than Observables.

[SX L <han...@ho...>]

Hi Eric,

Thanks for looking into this issue. Yes, I identified the issue in my windows x64 debug version. The issue could be hard to reproduce since my release version gives me normal exit code. It really depends on the sequence of calling dtors of those shared pointers at exit.

So I would suggest moving away from shared pointers in Singletons. Create a singltonDestroyer or use static instances.

Get Outlook for Android<https://aka.ms/ghei36>

________________________________
From: Eric Ehlers <eri...@re...>
Sent: Monday, November 9, 2020, 11:42 PM
To: SX L
Cc: qua...@li...
Subject: Re: [Quantlib-users] ObservableSettings is deleted early than Observables.


Hi Philip,

I tried unsuccessfully to recreate the error.  You did not say which operating system you are using but you mention "debug version" which sounds like visual studio so I tested it on Windows - I'm using Visual Studio 2019 on 64-bit WIndows 10.  I built it using "Debug (static runtime)" x64 and I ran ClientCppDemo and it behaves as I expected (no access violation).  Anything else I can try to recreate the error?

I have not examined the code but your remarks sound sensible to me.  What resolution would you suggest?

Regards,
Eric

On 10/7/20 6:43 AM, SX L wrote:
I was trying the Quantlib Addin. The QLADemo exits abnormally in the debug version with an access violation exception. It turned out to be the case that ObervableSettings is deleted earlier than some observer and observables at exit. So in the dtor of the observer, ObservableSettings is called to check if update is deferred, hence causing the exception. I feel it's unsafe to use shared pointers inside a singleton. Any suggestion? Thanks.

Regards,

Philip


Get Outlook for Android<https://aka.ms/ghei36>




_______________________________________________
QuantLib-users mailing list
Qua...@li...<mailto:Qua...@li...>
https://lists.sourceforge.net/lists/listinfo/quantlib-users