More on Compatibility between Series and xAct
Geoffrey Compere
Thanks a lot for all bug fixing ! All previous bugs are over.
Now, the Series is not yet over, if I can joke about it :).
I encountered a new problem between IsIndexOf : here it is:
IsIndexOf::nouse: Attempting to apply IsIndexOf on
Integrate`ExpandHold[-h^(1) c$10714$10800$12258 c
$10714$10800$12258
h^(0) c$10718$10800$12258 d$10718$10800$12258].
This is similar to the previous bug. If I only have
Unprotect[IsIndexOf];
IsIndexOf[expr_, index_, f_] :=
With[{iios = IsIndexOf[#, index, f] & /@ expr[[3]]},
If[Not[SameQ @@ iios],
Throw[Message[IsIndexOf::error,
"SeriesData with inconsistent index structure."]; ERROR[expr]],
And @@ iios]] /; Head[expr] === SeriesData;
IsIndexOf[Integrate`ExpandHold[_Power], _, _] := IndexList[];
Protect[IsIndexOf];
It is not sufficient to solve the problem.
By the way, it seems that IndexSet always adds new set of numbers to
dummy indices. If I continue to build up tensors of tensors of tens...
there might be an issue with dealing with indices of very large size,
isn't ?
Thanks a lot for your patience !
Geoffrey
JMM
> Thanks a lot for all bug fixing ! All previous bugs are over.
They were not bugs! xAct has never tried to work on SeriesData
expressions before. We are extending it now.
> I encountered a new problem between IsIndexOf [...]
OK, so this is an issue with more general types of expressions. Then I
suggest this more agressive solution (this already includes the
definition for Power):
IsIndexOf[ Integrate`ExpandHold[ expr_ ], index_, f_ ] :=
IsIndexOf[ expr, index, f ];
If this still gave you problems, remove those definitions and try this
other trick:
DefInertHead[ Integrate`ExpandHold ]
> By the way, it seems that IndexSet always adds new set of numbers to
> dummy indices. If I continue to build up tensors of tensors of tens...
> there might be an issue with dealing with indices of very large size,
> isn't ?
I don't think so. They are still symbols, but with a long name. If you
don't like that, use either ReplaceDummies or ScreenDollarIndices to
change them to "nicer" indices. The recommended way is working with
$PrePrint = ScreenDollarIndices, which hides those indices anyway...
Cheers,
Jose.
Geoffrey Compere
Indeed, you are extending the realm of xTensor, sorry for the
misleading statement.
It seems that the definition of IsIndexOf as
IsIndexOf[ Integrate`ExpandHold[ expr_ ], index_, f_ ] :=
IsIndexOf[ expr, index, f ];
solves the problem ! Thanks,
Geoffrey
Geoffrey Compere
I am stuck on another error, which is independent of Series. Acting
with Simplication on a sum of tensors ( they depend on the parameter
p), I got this:
ToCanonical::noident: Unknown expression not canonicalized:
xAct`xTensor`Private`addTMPCovD[i^(1)a a$19925,ParamD[p],ParamD] .
Validate::inhom: Found inhomogeneous indices: {IndexList[a],
IndexList[-a$19925]}.
Throw::nocatch: Uncaught Throw[Null] returned to top level.
I think that there is something going on with ToCanonical, but I'm not
sure what..
Thanks for reading this email !
Geoffrey
On Nov 24, 3:10 am, JMM <j...@xact.es> wrote:
Alfonso García-Parrado
> I am stuck on another error, which is independent of Series. Acting
> with Simplication on a sum of tensors ( they depend on the parameter
> p), I got this:
>
> ToCanonical::noident: Unknown expression not canonicalized:
> xAct`xTensor`Private`addTMPCovD[i^(1)a a$19925,ParamD[p],ParamD] .
> Validate::inhom: Found inhomogeneous indices: {IndexList[a],
> IndexList[-a$19925]}.
> Throw::nocatch: Uncaught Throw[Null] returned to top level.
>
could you send the actual expression which gave you this error when
acting with Simplification ?
Thank you.
Alfonso.
Geoffrey Compere
Thanks for asking. I looked at the simplest example and found it:
ParamD[PhaseSpacePoint][i1[a, b]]*PD[-b][\[Sigma][]] //
Simplification
gives the error. I previously defined the scalar \[Sigma] and the
tensor i1[a,b] in the standard way as
DefTensor[\[Sigma][], {GeneralManifold, PhaseSpacePoint},
PrintAs -> "\[Sigma]"]
DefTensor[i1[-a, -b], {GeneralManifold, PhaseSpacePoint},
Symmetric[{-a, -b}], PrintAs -> "\!\(\*SuperscriptBox[\"i\",
RowBox[{\"(\", \"1\", \")\"}]]\)"]
It seems to be an internal error of xTensor
Best regards,
Geoffrey
On Nov 28, 1:50 pm, Alfonso García-Parrado <alfo...@math.uminho.pt>
wrote:
JMM
I cannot reproduce this problem. Please provide a complete set of
commands leading to the error message, or a small notebook. Note that
your original email mentioned Simplification acting on a sum of
tensors, but the example you provided has Simplification acting on a
product of tensors.
And please start a new thread when you report a new issue. It is
confusing to have a discussion on Simplification + ParamD in a thread
called "Series and xAct".
Cheers,
Jose.
On Nov 28, 9:08 am, Geoffrey Compere <geoffreycomp...@gmail.com>
wrote:
Geoffrey Compere
Here is a simple sample file where the error occurs.
It seems to be a clash between ToCanonical and ParamD.
Sorry for posting in the wrong thread, I was for a moment confused on the origin of the problem. I simplified the expression with the error from a complicated Series to a Plus to a Times to a very simple expression that I now put in the sample file.
Best,
Geoffrey
JMM
> Here is a simple sample file where the error occurs.
> It seems to be a clash between ToCanonical and ParamD.
Thanks. There is indeed missing code to handle a particular case
involving canonicalization of ParamD objects and metrics. I will add
this in the next version of xTensor. For the time being I recommend to
switch to alternative canonicalization methods. For example use the
option Method->Implode in ToCanonical or Simplification when you find
this problem. You can also use the detour ... // ToTensorDerivative //
Simplification // FromTensorDerivative.
Cheers,
Jose.
