How count NaN values and replace NaNs in a sequence with that number?

1 vue (au cours des 30 derniers jours)
Mr M.
Mr M. le 30 Juin 2018
Modifié(e) : Stephen23 le 30 Juin 2018
I have a vector [0 0 NaN 0 0 NaN NaN NaN 0 0 0 0 0 NaN] and I want [0 0 1 0 0 3 3 3 0 0 0 0 0 1]. Is it possible without a for cycle?

Réponse acceptée

Guillaume
Guillaume le 30 Juin 2018
Modifié(e) : Guillaume le 30 Juin 2018
I don't think it's entirely possible to do it without a loop, but the loop can be limited to the number of sequences (3 steps in your example)
v = [0 0 NaN 0 0 NaN NaN NaN 0 0 0 0 0 NaN];
seqedges = find(diff([false, isnan(v), false]));
seqstarts = seqedges(1:2:end);
seqends = seqedges(2:2:end);
seqlengths = seqends - seqstarts;
for seq = 1:numel(seqstarts)
v(seqstarts(seq):seqends(seq)-1) = seqlengths(seq);
end
edit: actually, it's possible without a loop:
v = [0 0 NaN 0 0 NaN NaN NaN 0 0 0 0 0 NaN];
nanlocs = isnan(v);
seqedges = find(diff([false, nanlocs, false]));
seqlengths = seqedges(2:2:end) - seqedges(1:2:end);
v(nanlocs) = repelem(seqlengths, seqlengths)

Plus de réponses (2)

Stephen23
Stephen23 le 30 Juin 2018
Modifié(e) : Stephen23 le 30 Juin 2018
No loops, no third-party functions, and works for all MATLAB versions:
>> V = [0 0 NaN 0 0 NaN NaN NaN 0 0 0 0 0 NaN];
>> X = isnan(V);
>> Y = cumsum(diff([0,X])>0);
>> D = diff([0,X,0]);
>> L = find(D<0)-find(D>0);
>> V(X) = L(Y(X))
V = 0 0 1 0 0 3 3 3 0 0 0 0 0 1

Jan
Jan le 30 Juin 2018
Modifié(e) : Jan le 30 Juin 2018
With FEX: RunLength (link):
a = [0 0 NaN 0 0 NaN NaN NaN 0 0 0 0 0 NaN];
[b, n] = RunLength(isnan(a));
Result = RunLength(b .* n, n);
The C-Mex version runs very fast, but you have to compile it at first. But you can use the M-version included in this submission also: RunLength_M

Catégories

En savoir plus sur Logical dans Help Center et File Exchange

Tags

Community Treasure Hunt

Find the treasures in MATLAB Central and discover how the community can help you!

Start Hunting!

Translated by