# Bit matrix of non-NaNs
**URL:** https://discourse.julialang.org/t/bit-matrix-of-non-nans/128376
**Category:** New to Julia
**Created:** [April 24, 2025, 3:24pm UTC](https://discourse.julialang.org/t/bit-matrix-of-non-nans/128376 "2025-04-24T15:24:06Z")
**Posts on this page:** 9
**Page:** 1
### Author:  [@Boris](https://discourse.julialang.org/u/Boris)
#### Post date: [April 24, 2025, 3:24pm UTC](https://discourse.julialang.org/t/bit-matrix-of-non-nans/128376/1 "2025-04-24T15:24:06Z")
Hi. If I have a matrix with NaNs I can find them with `isnan.()`. Sorry for the stupid question, but how do I get the non-NaN bit matrix?
suppose I have
```julia
YY = rand(10,3); YY[2,1] = NaN; YY[4,2] = NaN; YY[3,3] = NaN;
Sm = isnan.(YY)
```
how do I get the “other” matrix where i have only the non-NaN values?
I tried
```julia
So = !isnan.(YY)
```
but that doesn’t work. I would do this in Matlab (with `~`) and somehow I expected this to work because I have seen similar synthax in Julia, for example with findall for Cartesian Indices
```julia
CI_nan =findall(isnan,YYt) # works
CI_non = findall(!isnan,YYt) # also works
```
but I don’t seem to find the correct syntax for the bit matrix…
Thanks.
---
### Author:  [@mcabbott](https://discourse.julialang.org/u/mcabbott)
#### Post date: [April 24, 2025, 3:30pm UTC](https://discourse.julialang.org/t/bit-matrix-of-non-nans/128376/2 "2025-04-24T15:30:20Z")
You need one more dot…
```julia
julia> @. !isnan(YY) # macro adds dots everywhere
10×3 BitMatrix:
1 1 1
0 1 1
...
julia> .!isnan.(YY) # infix operator takes . before, like .*
10×3 BitMatrix:
1 1 1
...
julia> (!isnan).(YY) # uses !(::Function)
10×3 BitMatrix:
1 1 1
...
julia> typeof(!isnan)
ComposedFunction{typeof(!), typeof(isnan)}
julia> Meta.@lower !isnan.(YY) # failed case, broadcasts then applies ! to array
:($(Expr(:thunk, CodeInfo(
@ none within `top-level scope`
1 ─ %1 = !
│ %2 = Base.broadcasted(isnan, YY)
│ %3 = Base.materialize(%2)
│ %4 = (%1)(%3)
└── return %4
))))
julia> Meta.@lower .!isnan.(YY) # one fused broadcast
:($(Expr(:thunk, CodeInfo(
@ none within `top-level scope`
1 ─ %1 = !
│ %2 = Base.broadcasted(isnan, YY)
│ %3 = Base.broadcasted(%1, %2)
│ %4 = Base.materialize(%3)
└── return %4
))))
```
---
### Author:  [@Boris](https://discourse.julialang.org/u/Boris)
#### Post date: [April 24, 2025, 3:39pm UTC](https://discourse.julialang.org/t/bit-matrix-of-non-nans/128376/3 "2025-04-24T15:39:01Z")
Missed that!
Thanks!
---
### Author:  [@DNF](https://discourse.julialang.org/u/DNF)
#### Post date: [April 24, 2025, 7:14pm UTC](https://discourse.julialang.org/t/bit-matrix-of-non-nans/128376/4 "2025-04-24T19:14:17Z")
Is it obvious why `.` should have higher precedence than `!`?
---
### Author:  [@Boris](https://discourse.julialang.org/u/Boris)
#### Post date: [April 25, 2025, 5:52am UTC](https://discourse.julialang.org/t/bit-matrix-of-non-nans/128376/5 "2025-04-25T05:52:10Z")
Hm, no I didn’t even think about that. Now that you mention it I did try `!.` yesterday but it was one of many tries didn’t think much about it.
That is a good learning experience, why?
---
### Author:  [@DNF](https://discourse.julialang.org/u/DNF)
#### Post date: [April 25, 2025, 6:08am UTC](https://discourse.julialang.org/t/bit-matrix-of-non-nans/128376/6 "2025-04-25T06:08:51Z")
What I meant was that it would be possible to interpret (parse) `!isnan.()` as `(!isnan).()`, but it is actually parsed as `!(isnan.())`. I was wondering why that should be the case.
---
### Author:  [@jw3126](https://discourse.julialang.org/u/jw3126)
#### Post date: [April 25, 2025, 6:27am UTC](https://discourse.julialang.org/t/bit-matrix-of-non-nans/128376/7 "2025-04-25T06:27:35Z")
Dot is not only for broadcast but also for getproperty. Consider !geometry.isvisible there is only one practical precedence here.
---
### Author:  [@DNF](https://discourse.julialang.org/u/DNF)
#### Post date: [April 25, 2025, 6:51am UTC](https://discourse.julialang.org/t/bit-matrix-of-non-nans/128376/8 "2025-04-25T06:51:51Z")
> [@jw3126](#):
>
> Dot is not only for broadcast but also for getproperty.
Yeah, that was the part I was missing.
---
### Author:  [@eldee](https://discourse.julialang.org/u/eldee)
#### Post date: [April 25, 2025, 7:18pm UTC](https://discourse.julialang.org/t/bit-matrix-of-non-nans/128376/9 "2025-04-25T19:18:32Z")
> [@Boris](#):
>
> I would do this in Matlab (with `~`) and somehow I expected this to work
You could use `So = .~isnan.(YY)`.