1 | initial version |

Based on Shashank comment, but see extra questions:

```
def changevar_definite(f, eqn, newvar, oldvar, a, b):
old_assumptions= assumptions() # Keep current assumptions for later restore
assume(oldvar > a, oldvar < b) # This is shitty (Maxima pb), see below
h(oldvar)= solve(eqn, newvar)[0].rhs()
g(newvar)= solve(eqn, oldvar)[0].rhs()
dx = diff(g,newvar)
forget()
assume(old_assumptions) # Unfortunately, extra assumptions don't stay local
return f.substitute(oldvar==g(newvar))*dx, h(a), h(b)
```

Okay, let's try it

```
integrand=changevar_definite(x*cos(x^2+1), u==x^2+1, u, x, 0, 2)
integrand # Output: (u |--> 1/2*cos(u), 1, 5)
```

Eventually

```
integral(integrand[0](u), u, integrand[1], integrand[2]) # -1/2*sin(1) + 1/2*sin(5)
```

But the following also works:

```
changevar_definite(x*cos(x^2+1), x==sqrt(u-1), u, x, 0, 2)
```

Now, the latter call breaks if I have `assume(oldvar >= a, oldvar <= b)`

instead as I did earlier, because Maxima asks about x being positive or zero. In fact, the assumptions mechanism is too sticky, what I change locally has global side effect. A worse problem is I can't force a more stringent assumption locally (say going from x >= 0 to x > 0).

1- How to efficiently handle those assumptions?

2- How to pass extra-assumptions as arguments to my `changevar_definite`

function?

2 | No.2 Revision |

Based on Shashank comment, but see extra ~~questions:~~questions below:

```
def changevar_definite(f, eqn, newvar, oldvar, a, b):
old_assumptions= assumptions() # Keep current assumptions for later restore
assume(oldvar > a, oldvar < b) # This is shitty (Maxima pb), see below
h(oldvar)= solve(eqn, newvar)[0].rhs()
g(newvar)= solve(eqn, oldvar)[0].rhs()
dx = diff(g,newvar)
forget()
assume(old_assumptions) # Unfortunately, extra assumptions don't stay local
return f.substitute(oldvar==g(newvar))*dx, h(a), h(b)
```

Okay, let's try it

```
integrand=changevar_definite(x*cos(x^2+1), u==x^2+1, u, x, 0, 2)
integrand # Output: (u |--> 1/2*cos(u), 1, 5)
```

Eventually

`integral(integrand[0](u), u, integrand[1], `~~integrand[2]) # ~~integrand[2])# -1/2*sin(1) + 1/2*sin(5)

But the following also works:

```
changevar_definite(x*cos(x^2+1), x==sqrt(u-1), u, x, 0, 2)
```

Now, the latter call breaks if I have `assume(oldvar >= a, oldvar <= b)`

instead as I did earlier, because Maxima asks about x being positive or zero. In fact, the assumptions mechanism is too sticky, what I change locally has global side effect. A worse problem is I can't force a more stringent assumption locally (say going from x >= 0 to x > 0).

1- How to efficiently handle those assumptions?

2- How to pass extra-assumptions as arguments to my `changevar_definite`

function?

3 | No.3 Revision |

Based on Shashank comment, but see extra questions below:

```
def changevar_definite(f, eqn, newvar, oldvar, a, b):
old_assumptions= assumptions() # Keep current assumptions for later restore
assume(oldvar > a, oldvar < b) # This is shitty (Maxima pb), see below
h(oldvar)= solve(eqn, newvar)[0].rhs()
g(newvar)= solve(eqn, oldvar)[0].rhs()
dx =
```~~diff(g,newvar)
~~diff(g, newvar)
forget()
assume(old_assumptions) # Unfortunately, extra assumptions don't stay local
return f.substitute(oldvar==g(newvar))*dx, h(a), h(b)

Okay, let's try it

```
integrand=changevar_definite(x*cos(x^2+1), u==x^2+1, u, x, 0, 2)
integrand # Output: (u |--> 1/2*cos(u), 1, 5)
```

Eventually

```
integral(integrand[0](u), u, integrand[1], integrand[2])# -1/2*sin(1) + 1/2*sin(5)
```

But the following also works:

```
changevar_definite(x*cos(x^2+1), x==sqrt(u-1), u, x, 0, 2)
```

Now, the latter call breaks if I have `assume(oldvar >= a, oldvar <= b)`

instead as I ~~did ~~had earlier, because Maxima asks about x being positive or zero. In fact, the assumptions mechanism is too sticky, what I change locally has global side effect. A worse problem is I can't force a more stringent assumption locally (say going from x >= 0 to x > 0).

1- How to efficiently handle those assumptions?

2- How to pass extra-assumptions as arguments to my `changevar_definite`

function?

4 | No.4 Revision |

Based on Shashank comment, but see extra questions below:

```
def changevar_definite(f, eqn, newvar, oldvar, a, b):
old_assumptions= assumptions() # Keep current assumptions for later restore
assume(oldvar > a, oldvar < b) # This is
```~~shitty ~~bad (Maxima pb), see below
h(oldvar)= solve(eqn, newvar)[0].rhs()
g(newvar)= solve(eqn, oldvar)[0].rhs()
dx = diff(g, newvar)
forget()
assume(old_assumptions) # Unfortunately, extra assumptions don't stay local
return f.substitute(oldvar==g(newvar))*dx, h(a), h(b)

Okay, let's try it

```
integrand=changevar_definite(x*cos(x^2+1), u==x^2+1, u, x, 0, 2)
integrand # Output: (u |--> 1/2*cos(u), 1, 5)
```

Eventually

```
integral(integrand[0](u), u, integrand[1], integrand[2])# -1/2*sin(1) + 1/2*sin(5)
```

But the following also works:

```
changevar_definite(x*cos(x^2+1), x==sqrt(u-1), u, x, 0, 2)
```

Now, the latter call breaks if I have `assume(oldvar >= a, oldvar <= b)`

instead as I had earlier, because Maxima asks about x being positive or zero. In fact, the assumptions mechanism is too sticky, what I change locally has global side effect. A worse problem is I can't force a more stringent assumption locally (say going from x >= 0 to x > 0).

1- How to efficiently handle those assumptions?

2- How to pass extra-assumptions as arguments to my `changevar_definite`

function?

Copyright Sage, 2010. Some rights reserved under creative commons license. Content on this site is licensed under a Creative Commons Attribution Share Alike 3.0 license.