fmpz_mat_snf_diagonal (and Nemo is_diagonal (not used))

[Grégory Vanuxem]

Hi,

I am writing an interface to Julia for FriCAS, and so also parts of
Nemo. And I wonder if this is the behavior you want, that is,
fmpz_mat_snf_diagonal does not check if the entry matrix is diagonal.
That is not a problem for me I will just code something like:

smithDiagonal(m) ==
diagonal?(m) => junfunc("snf_diagonal", m)
error "smithDiagonal: not a diagonal matrix"

Greg

================================================================

Nemo:

snf_diagonal(x::ZZMatrix)

Given a diagonal matrix $x$ compute the Smith normal form of $x$.
"""
function snf_diagonal(x::ZZMatrix)
z = similar(x)
@ccall libflint.fmpz_mat_snf_diagonal(z::Ref{ZZMatrix},
x::Ref{ZZMatrix})::Nothing
return z
end

jlFriCAS:

a:=matrix([[1,4],[7,9]])

+1 4+
(2) | |
+7 9+

Type: Matrix(Integer)
(3) -> a:=matrix([[1,4],[7,9]])@JLMatrix(NMInteger)

(3)

[1 4]
[7 9]

Type: JLMatrix(NMInteger)

(4) -> smith a

(4)

[1 0]
[0 19]

Type: JLMatrix(NMInteger)
(5) -> smithDiagonal a

(5)

[1 4]
[7 9]

Type: JLMatrix(NMInteger)
(6) -> diagonal? a

(6) false

Type: Boolean
(7) -> jlType a

(7) ZZMatrix

Type: Symbol

[Edgar Costa]

That is the intent, in general ,the user is the one responsible to provide valid input. 

--

---
You received this message because you are subscribed to the Google Groups "flint-devel" group.
To unsubscribe from this group and stop receiving emails from it, send an email to flint-devel...@googlegroups.com.
To view this discussion, visit https://groups.google.com/d/msgid/flint-devel/CAHnU2dax9W9dJt_84B7aUtXzC8PVO52PuoQio_xNx77TXTa6Sw%40mail.gmail.com.

[Albin Ahlbäck]

Hi Grégory,

I will read this as: "Is it expected that `fmpz_mat_snf_diagonal' does
not check that the matrix is indeed diagonal?"

Usually we do not check things if an argument is called (varies a bit
throughout FLINT, but in general this is a good strategy if it can be
easily checked beforehand). However, the docstring could reflect this
better.

Best,
Albin

[Grégory Vanuxem]

Le dim. 21 juin 2026 à 22:31, Edgar Costa <edgardi...@gmail.com> a écrit :
>
> That is the intent, in general ,the user is the one responsible to provide valid input.

This is what I guessed in fact, and finally I also used this strategy.
In (jl)FriCAS there are also some routines that do not perform checks,
this is usually stated in the documentation. This is what I did, as
evoked by Albin.

Greg

> To view this discussion, visit https://groups.google.com/d/msgid/flint-devel/CA%2BiQ7x4m1sQ3wquYSgiAiKDrV0mEv%2Bwc-HHdYCY0aivzwDVRow%40mail.gmail.com.

[Grégory Vanuxem]

Le dim. 21 juin 2026 à 22:32, Albin Ahlbäck <albin....@gmail.com> a écrit :
>
> Hi Grégory,
>
> I will read this as: "Is it expected that `fmpz_mat_snf_diagonal' does
> not check that the matrix is indeed diagonal?"

Right!

> Usually we do not check things if an argument is called (varies a bit
> throughout FLINT, but in general this is a good strategy if it can be
> easily checked beforehand). However, the docstring could reflect this
> better.
>

I also did that. And about performance, FLINT is intended to be highly
efficient and it is, I also think that here it is up to the user to
know what it is doing. Checking everything would be counterproductive.

Thanks for your responses.

Greg