How can I make GetFiles() exclude files with extensions that start with the search extension?

Question

I am using the following line to return specific files...

FileInfo file in nodeDirInfo.GetFiles("*.sbs", option)

But there are other files in the directory with the extension .sbsar, and it is getting them, too. How can I differentiate between .sbs and .sbsar in the search pattern?

You cannot (at least with GetFiles/GetDirectories). This is a "limitation" of the search pattern. You should iterate through the results and filter manually the ones you want. — varocarbas, Nov 27 '13 at 11:15
also take a look at this: http://msdn.microsoft.com/en-us/library/wz42302f(v=vs.110).aspx — Ahmed ilyas, Nov 27 '13 at 11:16

Daniel Peñalba · Answer 1 · 2013-11-27T12:34:02.977

9

The issue you're experiencing is a limitation of the search pattern, in the Win32 API.

A searchPattern with a file extension (for example *.txt) of exactly three characters returns files having an extension of three or more characters, where the first three characters match the file extension specified in the searchPattern.

My solution is to manually filter the results, using Linq:

nodeDirInfo.GetFiles("*.sbs", option).Where(s => s.EndsWith(".sbs"),
    StringComparison.InvariantCultureIgnoreCase));

edited Nov 27 '13 at 12:34

answered Nov 27 '13 at 11:18

Daniel Peñalba

30,507
32
137
219

2

You don't account for letter case here. – David Heffernan Nov 27 '13 at 11:51

score 7 · Accepted Answer · answered Nov 27 '13 at 11:30

7

Try this, filtered using file extension.

  FileInfo[] files = nodeDirInfo.GetFiles("*", SearchOption.TopDirectoryOnly).
            Where(f=>f.Extension==".sbs").ToArray<FileInfo>();

answered Nov 27 '13 at 11:30

Sid

197
12

David Heffernan · Answer 3 · 2013-11-27T12:11:00.163

6

That's the behaviour of the Win32 API (FindFirstFile) that is underneath GetFiles() being reflected on to you.

You'll need to do your own filtering if you must use GetFiles(). For instance:

GetFiles("*", searchOption).Where(s => s.EndsWith(".sbs", 
    StringComparison.InvariantCultureIgnoreCase));

Or more efficiently:

EnumerateFiles("*", searchOption).Where(s => s.EndsWith(".sbs", 
    StringComparison.InvariantCultureIgnoreCase));

Note that I use StringComparison.InvariantCultureIgnoreCase to deal with the fact that Windows file names are case-insensitive.

If performance is an issue, that is if the search has to process directories with large numbers of files, then it is more efficient to perform the filtering twice: once in the call to GetFiles or EnumerateFiles, and once to clean up the unwanted file names. For example:

GetFiles("*.sbs", searchOption).Where(s => s.EndsWith(".sbs", 
    StringComparison.InvariantCultureIgnoreCase));
EnumerateFiles("*.sbs", searchOption).Where(s => s.EndsWith(".sbs", 
    StringComparison.InvariantCultureIgnoreCase));

edited Nov 27 '13 at 12:11

answered Nov 27 '13 at 11:17

David Heffernan

601,492
42
1,072
1,490

@Joey That just feels a little dirty to me, duplicating the filter. But perhaps it would have a perf implication. If not then I'd rather have just the one filter. – David Heffernan Nov 27 '13 at 11:21
It's faster, though ;-) In my small test here (running over our complete source folder, searching for `*.cpp`) it's about 10–25 % faster to specify the filter in `GetFiles` too. `EnumerateFiles` is slightly slower, but probably uses much less memory, especially for large result sets. – Joey Nov 27 '13 at 11:36
@Joey Yes, I think that's reasonable. I guess it comes down to a balance between perf and purity! I've covered this in the answer now. – David Heffernan Nov 27 '13 at 11:51
1

Oh, and I guess `.EndsWith(".sbs", StringComparison.InvariantCultureIgnoreCase)` would be a better option that's also resistant to culture, as the file system ignores the culture for its case-insensitivity. – Joey Nov 27 '13 at 12:08
@Joey Thanks. Showing my ignorance with `ToLower()`! – David Heffernan Nov 27 '13 at 12:11

Anirudha · Answer 4 · 2013-11-27T11:26:16.470

1

Its mentioned in docs

When using the asterisk wildcard character in a searchPattern,a searchPattern with a file extension of exactly three characters returns files having an extension of three or more characters.When using the question mark wildcard character, this method returns only files that match the specified file extension.

edited Nov 27 '13 at 11:26

answered Nov 27 '13 at 11:20

Anirudha

32,393
7
68
89

1

It would be marvellous if this would be true. Unfortunately, it is not and and here comes a new episode of the poor descriptions of searchPattern in MSDN :) I felt curious and did some tests and here come my conclusions... – varocarbas Nov 27 '13 at 11:29
@varocarbas indeed..wonder where to use `?`.OP can use `*a?.sbs`..Though that would require `a` to be somewhere in the file name – Anirudha Nov 27 '13 at 11:32
3

`nodeDirInfo.GetFiles("5?.txt");` returns any file with just .txt (not .txtwhatever) containing two characters in the name, one of them being a 5. `nodeDirInfo.GetFiles("?.txt");` Returns any .txt file with just one character in its name (not including .txtwhatever). You can get only *.txt by using a `????.txt` approach if you know the maximum length of the file names you are looking for (??.txt retuns all the files with 1 or 2 characters in its name; ???.txt all the ones with 1,2 and 3, etc.). – varocarbas Nov 27 '13 at 11:40
this was the answer I was hoping would work. But '?.sbs' returned nothing and '*?.sbs" returned all files with 'sbs' in the extension. The only thing these file names have in common is the extension. I imagine that would be the case with many such searches. I agree with varocarbas, the docs are not clear. – topofsteel Nov 27 '13 at 21:45

How can I make GetFiles() exclude files with extensions that start with the search extension?

4 Answers4