QuantLib › quantlib-users

Behavior of (ql)TimeSeries

Classic

List

Threaded

4 messages Options

Ballabio Gerardo-4

Behavior of (ql)TimeSeries

Hi all,
I'm exploring time series in QuantLibXL. I discovered an aspect of their
behavior that I find puzzling.

To start, I built a TimeSeries object by calling qlTimeSeries on a pair
of dates/values vectors. Then I called qlTimeSeriesValue on various
dates. Of course when a date belongs to the input dates set that I built
the TimeSeries from, the corresponding value is returned. When a date
doesn't belong to the input set, I was expecting to obtain the value on
the latest previous date, or maybe some kind of interpolated value.
Instead it always returns 3.40282E+38, which I suppose is an exceptional
value meaning "no valid value" (if I remember correctly it is the
largest possible value for 32-bit IEEE floats). As far as I can see
there is no way, at least in QuantLibXL, to change that behavior (if
there is one please tell me). I don't like it very much but I understand
that, as the most conservative choice, you can argue it's the right
thing to do.

Then came the real surprise. I discovered that, when I call
qlTimeSeriesValue on a nonexistent date, that date together with the
value of 3.40282E+38 _get added to the time series_!!! (The attached
excel file demonstrates that.)

I can't think of a good reason for doing that. If the function returned
an interpolated value, I see how caching it could give a performance
improvement, especially if the interpolation formula was complex. But
how is it useful to cache the value if it's always 3.40282E+38??? I
don't think there's any performance gain to be obtained from that.
Finding whether a date is already there takes almost the same time if
it's there as if it isn't (I'm assuming some kind of binary search is
used). In fact the performance might even get worse, because you have to
insert the new date/value pair, and the vectors might have to be
reallocated.

Instead I can think of at least one reason why it seems a bad idea:
logically qlTimeSeriesValue should be a read-only operation which leaves
the target object unmodified, but in fact it isn't. Caching the date
isn't just a hidden implementation detail, it can have side effects. For
example if I call it on a date that is outside the current range, the
range is extended, thus the return value of either qlTimeSeriesFirstDate
or qlTimeSeriesLastDate changes (this actually happened to me, that's
how I noticed). Or if I want to interpolate in my excel sheet between
available dates, I must be very careful to _never_ call
qlTimeSeriesValue on missing dates, or I'd spoil it.

Thanks
Gerardo

<!DOCTYPE html PUBLIC "-//W3C//DTD XHTML 1.0 Strict//EN" "http://www.w3.org/TR/xhtml1/DTD/xhtml1-strict.dtd">
<html xmlns="http://www.w3.org/1999/xhtml">
<head></head>
<body>
<div style="font-family:Calibri;font-size:10px">
Banca Profilo S.p.A.
Corso Italia, 49 - 20122 Milano - Tel. 02 58408.1, Fax 02 5831 6057
Capitale Sociale Euro 136.794.106,00 i.v.
Iscrizione al Registro Imprese di Milano, C.F. e P.IVA 09108700155 - [hidden email]
Iscritta all’Albo delle Banche e dei Gruppi bancari
Aderente al Fondo Interbancario di Tutela dei depositi
Aderente al Conciliatore Bancario Finanziario e all’Arbitro Bancario Finanziario
Appartenente al Gruppo bancario Banca Profilo e soggetta all’attività di direzione e coordinamento di Arepo BP S.p.A.

DISCLAIMER:
The information transmitted may contain confidential and/or privileged material.
Any review, retransmission, dissemination or other use of, or taking of any action in reliance upon,
this information by persons or entities other than the intended recipient is prohibited.
If you received this in error, please contact the sender and delete the material from any computer.
</div>
</body>
</html>

------------------------------------------------------------------------------
Live Security Virtual Conference
Exclusive live event will cover all the ways today's security and
threat landscape has changed and how IT managers can respond. Discussions
will include endpoint security, mobile security and the latest in malware
threats. http://www.accelacomm.com/jaw/sfrnl04242012/114/50122263/
_______________________________________________
QuantLib-users mailing list
[hidden email]
https://lists.sourceforge.net/lists/listinfo/quantlib-users

timeseries.xlsx (13K) Download Attachment

Luigi Ballabio

Re: Behavior of (ql)TimeSeries

On Thu, May 24, 2012 at 2:32 PM, Ballabio Gerardo
<[hidden email]> wrote:
> Hi all,
> I'm exploring time series in QuantLibXL. I discovered an aspect of their
> behavior that I find puzzling.
>
> When a date
> doesn't belong to the input set [...] it always returns 3.40282E+38, which I suppose is an exceptional
> value meaning "no valid value"

Yes, that's known as Null<double>() in the library. See <ql/utilities/null.hpp>.

> Then came the real surprise. I discovered that, when I call
> qlTimeSeriesValue on a nonexistent date, that date together with the
> value of 3.40282E+38 _get added to the time series_!!!

That's because of the behavior of the std::map<Date, double> underneath.
If you look at the implementation of TimeSeries in C++, you'll see
that the const version of operator[] returns Null without adding it,
but the non-const has to return a reference and therefore needs an
entry in the map to refer to. If you didn't add it, you couldn't
write "series[d] = v;" to add a new value.

The real question (but I'm not overly familiar with that part, so
we'll have to rely on someone else) might be: why qlTimeSeriesValue,
which should only read from the time series, uses the non-const
version of operator[] instead of the const version?

Luigi

------------------------------------------------------------------------------
Live Security Virtual Conference
Exclusive live event will cover all the ways today's security and
threat landscape has changed and how IT managers can respond. Discussions
will include endpoint security, mobile security and the latest in malware
threats. http://www.accelacomm.com/jaw/sfrnl04242012/114/50122263/
_______________________________________________
QuantLib-users mailing list
[hidden email]
https://lists.sourceforge.net/lists/listinfo/quantlib-users

Ballabio Gerardo-4

R: Behavior of (ql)TimeSeries

Da: Luigi Ballabio [mailto:[hidden email]]
> > Then came the real surprise. I discovered that, when I call
> > qlTimeSeriesValue on a nonexistent date, that date together with the
> > value of 3.40282E+38 _get added to the time series_!!!
>
> That's because of the behavior of the std::map<Date, double>
underneath.

> If you look at the implementation of TimeSeries in C++, you'll see
> that the const version of operator[] returns Null without adding it,
> but the non-const has to return a reference and therefore needs an
> entry in the map to refer to. If you didn't add it, you couldn't
> write "series[d] = v;" to add a new value.
>
> The real question (but I'm not overly familiar with that part, so
> we'll have to rely on someone else) might be: why qlTimeSeriesValue,
> which should only read from the time series, uses the non-const
> version of operator[] instead of the const version?

You're saying that wasn't intentional, aren't you?

Gerardo

<!DOCTYPE html PUBLIC "-//W3C//DTD XHTML 1.0 Strict//EN" "http://www.w3.org/TR/xhtml1/DTD/xhtml1-strict.dtd">
<html xmlns="http://www.w3.org/1999/xhtml">
<head></head>
<body>
<div style="font-family:Calibri;font-size:10px">
Banca Profilo S.p.A.
Corso Italia, 49 - 20122 Milano - Tel. 02 58408.1, Fax 02 5831 6057
Capitale Sociale Euro 136.794.106,00 i.v.
Iscrizione al Registro Imprese di Milano, C.F. e P.IVA 09108700155 - [hidden email]
Iscritta all’Albo delle Banche e dei Gruppi bancari
Aderente al Fondo Interbancario di Tutela dei depositi
Aderente al Conciliatore Bancario Finanziario e all’Arbitro Bancario Finanziario
Appartenente al Gruppo bancario Banca Profilo e soggetta all’attività di direzione e coordinamento di Arepo BP S.p.A.

DISCLAIMER:
The information transmitted may contain confidential and/or privileged material.
Any review, retransmission, dissemination or other use of, or taking of any action in reliance upon,
this information by persons or entities other than the intended recipient is prohibited.
If you received this in error, please contact the sender and delete the material from any computer.
</div>
</body>
</html>

------------------------------------------------------------------------------
Live Security Virtual Conference
Exclusive live event will cover all the ways today's security and
threat landscape has changed and how IT managers can respond. Discussions
will include endpoint security, mobile security and the latest in malware
threats. http://www.accelacomm.com/jaw/sfrnl04242012/114/50122263/
_______________________________________________
QuantLib-users mailing list
[hidden email]
https://lists.sourceforge.net/lists/listinfo/quantlib-users

Luigi Ballabio

Re: Behavior of (ql)TimeSeries

On Thu, May 24, 2012 at 3:57 PM, Ballabio Gerardo
<[hidden email]> wrote:
>> The real question (but I'm not overly familiar with that part, so
>> we'll have to rely on someone else) might be: why qlTimeSeriesValue,
>> which should only read from the time series, uses the non-const
>> version of operator[] instead of the const version?
>
> You're saying that wasn't intentional, aren't you?

It might be, but I'm not positive. There might be a reason that escapes me.

Luigi

------------------------------------------------------------------------------
Live Security Virtual Conference
Exclusive live event will cover all the ways today's security and
threat landscape has changed and how IT managers can respond. Discussions
will include endpoint security, mobile security and the latest in malware
threats. http://www.accelacomm.com/jaw/sfrnl04242012/114/50122263/
_______________________________________________
QuantLib-users mailing list
[hidden email]
https://lists.sourceforge.net/lists/listinfo/quantlib-users