Vectorizing unicode2native

Hi

I am trying to remove the non-compatible letter with windows-1252 encoding. I can do this using a for-loop, but trying to speed up this algorithm if possible.

events_Description =["1:BRAZIL-ANAPOLIS";"2:BRAZIL-Barra de São Francisco";"3:BRAZIL-CAJAMAR";"4:BRAZIL-CAMAÇARI";"5:BRAZIL-CANSANÇÃO";"6:BRAZIL-CONCEIÇÃO DE FEIRA";"7:BRAZIL-CONCEIÇÃO DO JACUÃPE"];
tdata = repmat("", height( events_Description), 1);
for runi= 1: length( events_Description)
    tnew = unicode2native( events_Description( runi), 'windows-1252');
    tnew( ismember( tnew, [129   141   143   144   157])) = [];
    tdata( runi)= native2unicode( tnew, 'windows-1252');
end

Answer 1

[Walter Roberson]

%precompute
L = false(1,MaximumPermittedCharacter) ;
L(native2unicode(1:255,'windows-1252'))  = true;
%if you want to disallow specific characters set them to false

%then the filtering becomes 
Message = Message(L(Message));

[Pete sherer]

sorry, i didn't quite follow what you are doing here.

What is the MaximumPermittedCharacter?

why using 1:255?

Message is basically the 'events_Description'?

[Walter Roberson]

You are creating a lookup table to determine whether each character is to be permitted or not. Then you just take the input and use it to index the table. But for that to be successful, the input characters from events_description cannot have any entries with character position larger than the size of your table.

If you know ahead of time that no input character will ever be beyond (say) U+23FF (which gets you smart quotes, infinity symbol, power-off-button character), then you can

MaximumPermittedCharacter = 0x23ff;
L = false(1,MaximumPermittedCharacter) ;
L(native2unicode(1:255,'windows-1252'))  = true;

to create the table.

The alternative is,

L = logical.empty;
L(native2unicode(1:255,'windows-1252'))  = true;
MaximumPermittedCharacter = size(L,2);

then the lookup becomes

Message(Message > MaximumPermittedCharacter) = [];
Message = Message(L(Message));

which requries two passes over the message instead of one pass

[Walter Roberson]

Message is basically the 'events_Description'?

Message would be events_Description{runi} -- a character vector