The guy sitting in front of my computer introduced a bug due to passing a NaN where an integer should have been.
And that NaN mapped to a -1 which was kind of hard to track down.

It would be nice to have a fixed way of converting from special double values to integers.
Currently that seems a bit unspecified.

Function DoStuff()
int a
int64 b
uint64 c
print"Mapping NaN to int"
a = NaN
b = NaN
c = NaNprintf"%x, ,%x, %x\r", a, b, c
Make/FREE/B a1 = {NaN}Make/FREE/B/U a2 = {NaN}Make/FREE/W b1 = {NaN}Make/FREE/W/U b2 = {NaN}Make/FREE/I c1 = {NaN}Make/FREE/I/U c2 = {NaN}Make/FREE/L d1 = {NaN}Make/FREE/L/U d2 = {NaN}print/D a1[0], a2[0], b1[0], b2[0], c1[0], c2[0], d1[0], d2[0]print"Mapping inf to int"
a = Inf
b = Inf
c = Infprintf"%x, ,%x, %x\r", a, b, c
Make/FREE/B a1 = {Inf}Make/FREE/B/U a2 = {Inf}Make/FREE/W b1 = {Inf}Make/FREE/W/U b2 = {Inf}Make/FREE/I c1 = {Inf}Make/FREE/I/U c2 = {Inf}Make/FREE/L d1 = {Inf}Make/FREE/L/U d2 = {Inf}print/D a1[0], a2[0], b1[0], b2[0], c1[0], c2[0], d1[0], d2[0]print"Mapping -inf to int"
a = -inf
b = -inf
c = -infprintf"%x, ,%x, %x\r", a, b, c
Make/FREE/B a1 = {-inf}Make/FREE/B/U a2 = {-inf}Make/FREE/W b1 = {-inf}Make/FREE/W/U b2 = {-inf}Make/FREE/I c1 = {-inf}Make/FREE/I/U c2 = {-inf}Make/FREE/L d1 = {-inf}Make/FREE/L/U d2 = {-inf}print/D a1[0], a2[0], b1[0], b2[0], c1[0], c2[0], d1[0], d2[0]End

•dostuff()
Mapping NaN to int
8000000000000000, ,8000000000000000, 8000000000000000
-1 255 -1 65535 -1 4294967295 -9223372036854775808 -9223372036854775808
Mapping inf to int
8000000000000000, ,8000000000000000, 8000000000000000
-1 255 -1 65535 -1 4294967295 -9223372036854775808 -9223372036854775808
Mapping -inf to int
8000000000000000, ,8000000000000000, 8000000000000000
0 0 0 0 0 0 -9223372036854775808 -9223372036854775808

IMHO the best solution would be to throw a RTE with rtFunctionErrors=1 but I would expect that this is too costly performance-wise.

Joined: 2009-10-07

Location: Germany

The guy sitting in front of my computer introduced a bug due to passing a NaN where an integer should have been.

And that NaN mapped to a -1 which was kind of hard to track down.

It would be nice to have a fixed way of converting from special double values to integers.

Currently that seems a bit unspecified.

•dostuff()

Mapping NaN to int

8000000000000000, ,8000000000000000, 8000000000000000

-1 255 -1 65535 -1 4294967295 -9223372036854775808 -9223372036854775808

Mapping inf to int

8000000000000000, ,8000000000000000, 8000000000000000

-1 255 -1 65535 -1 4294967295 -9223372036854775808 -9223372036854775808

Mapping -inf to int

8000000000000000, ,8000000000000000, 8000000000000000

0 0 0 0 0 0 -9223372036854775808 -9223372036854775808

IMHO the best solution would be to throw a RTE with rtFunctionErrors=1 but I would expect that this is too costly performance-wise.