The comment from @Jehan deserves to be explained better in the Manual. This
took a long time to debug. The problem is that `cast[]` is not simply "unsafe";
it's downright deceitful.
I had an if-statement which was never evaluating to `true`. I couldn't believe
it, so I printed the elements of the expression, and then the full expression.
It was always printed as `true`! But `if` looked at the same expression
differently than `echo` did.
Here it is:
if aln.aln_str_size > 500 and
(cast[cdouble](aln.dist) / cast[cdouble](aln.aln_str_size)) <
max_diff:
echo "HELLO"
See if you can figure it out...
...
...
...
...
For the answer, I had to look at the generated source code. (I am so thankful
to be able to view the generated code.) Because `dist` and `size` are used as
both actual type (`int`) and casted type (`double`), the **Nim** compiler
decided to use a union:
union { int source; double dest; } LOC34;
But in my `echo` statements, there is no union, so a C-style cast is performed
instead.
Why not just be clear in the Manual:
> "cast[]" assumes that the types are bitwise compatible, e.g. "int32" and
> "uint32". Beware "cast[]" in general. For example, never use "cast[double]"
> to convert from an integer, nor vice versa.
I made this mistake because I am cleaning up thousands of lines of code
produced by **c2nim**, and it's hard to catch everything. It started as this:
if (aln->aln_str_size > 500 &&
((double) aln->dist / (double) aln->aln_str_size) < max_diff) {
