有什么方法可以加速必须在20000+行中搜索16x字符串的Matlab代码吗？

Question

Basically the way the code works is that the user types in a string (in my case a timestamp like 7:29:29 AM - 2:33:33 PM) and the code reads in a data file from excel that has those strings in it and all the data: filename = get(handles.File,'String'); [Data,Text] = xlsread(filename,2);

IndexStart=strmatch(get(handles.StartTime,'String'),Text,'exact'); %start time
IndexEnd=strmatch(get(handles.EndTime,'String'),Text,'exact'); %end time

seconds = IndexEnd-IndexStart;
PlotData = Data([IndexStart: IndexEnd],:);

It then searches for the row number of that time stamp in Text and copies the corresponding data section from the data for this time range so that I can plot it. This data is collected for 8+ hours at 1 sample/sec so there are easily 30000 rows in the excel file to search through. This large chunk of data is going to be plotted with labels on the plot for different events (assuming they put an event for every box but i take that into account with the if statement). The way I have this set up now is in a gui where the user places in timestamp values as strings and the code searches for them:

if isempty(get(handles.Task16End,'String')) 
IndexTextTask16End  = IndexStart;
else
IndexTextTask16End=strmatch(get(handles.Task16End,'String'),Text,'exact'); %row location for timestamp
end
Task16Span=IndexTextTask16End-IndexTextTask15End; %timespan of this event
Task16LineLocation=Task15LineLocation+ Task16Span/3600; %location for vertical line on graph

So i have up to 16 tasks that can be inputted which means that the program has to search through EVERY single available cell in a matlab matrix for these strings until it runs through the code. How can I do this more efficiently? maybe set it to search until it finds a truly empty cell? That would at least limit my search to the given data instead of the entire array possible.

Answer 1

If you're reading an Excel spreadsheet for start/end times and doing all this other work in MATLAB, consider converting the times to to their datenum representation after your xlsread . This way you can compare numbers, not strings - much faster. Given this, you could use logical indexing to build your desired data:

times   = datenum(Text); % assumed Text is just a cell array of times
t_start = datenum(get(handles.StartTime,'String'));
t_end   = datenum(get(handles.EndTime, 'String'));

plotData = Data(times >= t_start & times <= t_end); % note the single &, which is different than &&

Answer 2

strcmp is often much faster than strmatch ; I tried it out and it was much faster on my system, don't know exactly why (like 1000x faster, I wasn't expecting that much difference).

It returns slightly different information - strcmp returns a logical array with 1 wherever there's a match - so to get the same output as with strmatch just wrap it with a find :

IndexStart=find(strcmp(get(handles.StartTime,'String'),Text)); %start time