Do you think this is something I could work on and help out with? Is it of interest to the maintainers? If so I could create a GH issue about it.
As an alternavice could anyone suggest a nice dynamical systems focused libraries based on sympy that could improve on this? On Tuesday, February 11, 2025 at 12:14:40 PM UTC+8 Plamen Dimitrov wrote: > One more important observation here: sympy can in fact handle different > index symbols in the case of differentiation just fine > ``` > expr1 = x*y + sp.sin(x)**2 > der1ij = sp.diff(expr1, x.subs(i, j)) > # result: 2*sin(x[i])*cos(x[i])*KroneckerDelta(i, j) + KroneckerDelta(i, > j)*y[i] > ``` > as I guess it will assume any "x" symbol can still depend on "x_i" even if > we differentiate with respect to "x_j". In other words, it supports the > possibility of i=j while integration does not. This means that there is > lack of symmetry in expectation of supported functionality between > integration and differentiation, I assume mostly because the Kroenecker > delta approach is only possible for differentiation. I assume there is no > way we can achieve something similar for integration here? > On Tuesday, February 11, 2025 at 11:28:48 AM UTC+8 Plamen Dimitrov wrote: > >> > You have to explicitly tell that indexed variables can overlap and >> intersect. >> >> I see, I had similar thoughts for explicit splits but was hoping there >> might be some automation on the side of sympy for this way. Indeed I guess >> this could stretch too much what sympy should do for us and a split remains >> viable solution for simple enough cases. I believe there could be more >> complex equations out there where such explicit split might not be easily >> possible (in addition to slightly harmed readability from a shorter form of >> each equation) but have to reach this point first to confirm. In any case, >> thanks for the quick reply! >> >> On Saturday, February 1, 2025 at 5:13:24 AM UTC+8 [email protected] >> wrote: >> >>> You have to explicitly tell that indexed variables can overlap and >>> intersect. Maybe this code fetches you the desired result, >>> Try to split the sum into parts >>> >>> dF_dxi = x[i] + sp.Sum(x[j], (j, 1, i-1)) + sp.Sum(x[j], (j, i+1, C)) + >>> x[i] >>> #outpur: (Sum(x[j], (j, 1, i - 1)) + Sum(x[j], (j, i + 1, C)))*x[i] + >>> x[i]**2 >>> >>> On Tuesday, 28 January 2025 at 15:10:22 UTC+5:30 [email protected] wrote: >>> >>>> Hi, does anyone know how to properly integrate this indexed sum for an >>>> xi term that is also present in the sum indexed by j below? >>>> ``` >>>> i, j, C = sp.symbols('i, j, k, C', integer=True) >>>> x = sp.IndexedBase('x') >>>> dF_dxi = x[i] + sp.Sum(x[j], (j, 1, C)) >>>> Fxi = sp.integrate(dF_dxi, x[i]) >>>> # result: x[i]**2/2 + x[i]*Sum(x[j], (j, 1, C)) >>>> ``` >>>> It seems the integration does not realize that one of the xj terms in >>>> the sum is an xi since it assumes the indices don't ever intersect even >>>> though these are both defined as integers. >>>> >>>> >>>> -- You received this message because you are subscribed to the Google Groups "sympy" group. To unsubscribe from this group and stop receiving emails from it, send an email to [email protected]. To view this discussion visit https://groups.google.com/d/msgid/sympy/309d7b45-89f0-422c-b890-dc1c031c79ean%40googlegroups.com.
