Unique function cell array conflict

Matlab is giving me a seemingly conflicting error message. It says that the 'rows' flag is ignored since NewParts is a cell array, however it says that it cannot perform the unique function since NewParts is not a cell array. whos says that NewParts is a cell array. Are there any ways to resolve this conflict?

>> Classification
Warning: 'rows' flag is ignored for cell arrays. 
> In cell.unique at 31
  In Classification at 103
??? Error using ==> cell.unique at 45
Input must be a cell array of strings.

Error in ==> Classification at 103
UNIQUE_NewParts = unique(NewParts,'rows');
whos
NewParts                 2122x21            3330030  cell

Answer 1

[Sean de Wolski]

It's a cell array but not a cell array of strings.

unique({magic(2),magic(2),rand(2)})

Fail, your error message.

unique({'hello world','what''s up','hello world'})

pass.

[Brian]

All of the cells in the NewParts array are strings except for the NaN cells. Is there a way to have these NaN cells recognized as strings? Also, my goal is to list the unique rows of the NewParts cell array.

[Sean de Wolski]

idx = cellfun(@(c)all(isnan(c(:))),YourCell);

YourCell{idx} = 'nan'; %nan as a string

[Sean de Wolski]

For unique rows one way is to horizontally concatenate all of cells together since they're all strings. It would fail in the case of

{'hello ' 'worldTuesday'

'hello world' 'Tuesday'}

Though

[Brian]

I tried the "YourCell" code above, but got the error: ??? The right hand side of this assignment has too few values to satisfy the left hand side.

[Sean de Wolski]

That must be what you tried to insert it into. There is no left hand side in my above code :)

[Brian]

Isn't "YourCell{idx}" the left hand side of "YourCell{idx} = 'nan';"? For me, YourCell is a list from Excel with strings and NaNs.

[Sean de Wolski]

Oh yes (doh!), I wasn't seeing the hidden comments. Can you edit the original comments to include the exact code you're using ans a small data set, please?

[Brian]

YourCell = raw_BOM(:,3)

idx = cellfun(@(c)all(isnan(c(:))),YourCell);

YourCell{idx} = 'nan'; %nan as a string

UNIQUE_PARTS = char(unique(YourCell));

YourCell =

'PARTNUM'

[NaN]

'183-1695-010'

'183-1695-010'

'913-3067-150'

'913-3067-150'

'913-3060-080'

...

??? The right hand side of this assignment has too few values to satisfy the left hand side.

Error in ==> Classification at 30

YourCell{idx} = 'nan'; %nan as a string

[Sean de Wolski]

YourCell =...

{'PARTNUM'

[NaN]

'183-1695-010'

'183-1695-010'

'913-3067-150'

'913-3067-150'

'913-3060-080' };

idx = cellfun(@(c)all(isnan(c(:))),YourCell);

YourCell{idx} = 'nan'; %nan as a string

UNIQUE_PARTS = char(unique(YourCell));

UNIQUE_PARTS =

183-1695-010

913-3060-080

913-3067-150

PARTNUM

nan

No issues...

[Brian]

Is your [NaN] actually Not A Number? Or is it a string or something? The real problem is that it comes from a blank cell in Excel.

[Sean de Wolski]

yes, my nan is _actually_ a nan. I convert it to a string that says nan but it is no longer a nan, it's a string (so unique can be called)

Answer 2

[Brian]

Thanks for your help Sean de. Since the first column was driving the unique parts, I used this solution.

[B,I,J] = unique(NewParts(:,1));

UNIQUE_NewParts = NewParts(I,:)

Answer 3

[Jim Hokanson]

I just stumbled upon this weird behavior recently. See: http://www.mathworks.com/matlabcentral/answers/74169-unique-rows-cell-array-ignored

In summary, a warning is thrown saying that unique will not work with the rows option, but then it proceeds perform unique on all elements anyways, at which point in time it fails due to type inconsistencies.

For others finding this answer, I've submitted a more general FEX submission a few years ago to address this problem: 25917: Unique Rows for a Cell Array