List files in folder which match pattern

Question

I need to list files in directory which match some pattern. I tried playing with Directory.GetFiles, but don't fully get why it behaves in some way.

1) For example, this code:

            string[] dirs = Directory.GetFiles(@"c:\test\", "*t");

            foreach (string dir in dirs)
            {
                Debugger.Log(0, "", dir);
                Debugger.Log(0, "", "\n");

            }

outputs this:

c:\test\11.11.2007.txtGif
c:\test\12.1.1990.txt
c:\test\2.tGift
c:\test\2.txtGif
c:\test\test.txt
...others hidden

You can see some files end with f but were still returned by query, why?

2) Also, this:

                string[] dirs = Directory.GetFiles(@"c:\test\", "*.*.*.txt");


                foreach (string dir in dirs)
                {
                    Debugger.Log(0, "", dir);
                    Debugger.Log(0, "", "\n");

                }

returns this:

c:\test\1.1.1990.txt
c:\test\1.31.1990.txt
c:\test\12.1.1990.txt
c:\test\12.31.1990.txt

But according to the documentation (http://msdn.microsoft.com/en-us/library/07wt70x2(v=vs.110).aspx) I think it had to return also this file which is in the directory:

11.11.2007.txtGif

since extension (in the query string) is 3 letters long, but it didn't. why? (when query extension is 3 letters long, doc says it will return extensions which start with specified extensions too, e.g., see Remarks).

---

Am I the only one who finds these results strange?

Is there any other approach you would recommend for using when one wants to list files in folder which match certain pattern?

User in my case may arbitrarily type some pattern, and I don't want to rely on method which I am unsure about the result (like it happened with GetFiles).

Answer 1

https://learn.microsoft.com/en-us/dotnet/api/system.io.directoryinfo.getfiles?view=netframework-4.5

The above Microsoft documentation is wrong as usual, it says this code:

            DirectoryInfo di = new DirectoryInfo(@"C:\Users\tomfitz\Documents\ExampleDir");
            Console.WriteLine("No search pattern returns:");

            Console.WriteLine();

            Console.WriteLine("Search pattern *2* returns:");
            foreach (var fi in di.GetFiles("*2*"))
            {
                Console.WriteLine(fi.Name);
                Console.WriteLine(fi.Fullname); // this reveals the bug
            }

should return the following but it does not

It still matches against the whole file path not just the filename.

Search pattern *2* returns:
log2.txt
test2.txt

Answer 2

All of this behavior is exactly as described in the documentation you've linked. Here's an excerpt of the pertinent bits:

When you use the asterisk wildcard character in a searchPattern such as "*.txt", the number of characters in the specified extension affects the search as follows:

• If the specified extension is exactly three characters long, the method returns files with extensions that begin with the specified extension. For example, "*.xls" returns both "book.xls" and "book.xlsx".

• In all other cases, the method returns files that exactly match the specified extension. For example, "*.ai" returns "file.ai" but not "file.aif".

When you use the question mark wildcard character, this method returns only files that match the specified file extension. For example, given two files, "file1.txt" and "file1.txtother", in a directory, a search pattern of "file?.txt" returns just the first file, whereas a search pattern of "file*.txt" returns both files. NoteNote

Because this method checks against file names with both the 8.3 file name format and the long file name format, a search pattern similar to "1.txt" may return unexpected file names. For example, using a search pattern of "1.txt" returns "longfilename.txt" because the equivalent 8.3 file name format is "LONGFI~1.TXT".

http://msdn.microsoft.com/en-us/library/wz42302f%28v=vs.110%29.aspx

The last paragraph above clearly explains your results when searching for *t. You can see this by using the command dir C:\test /x to show the 8.3 filenames. Here, C:\test\11.11.2007.txtGif matches *t because its 8.3 filename is 111120~1.TXT.

For the treatment of *.*.*.txt, I think you're either mis-interpreting the first bit about three-letter file extensions or perhaps it wasn't written quite clearly. Note that they quite specifically mentioned wildcard usage 'in a searchPattern such as "*.txt"'. Your search pattern doesn't match that, so you have to read between the lines a bit to see why their comment about three-letter file extensions applies to the example they gave but not yours. Really, I think that whole top section can be ignored if you just put a bit of thought into the last bit about 8.3 filenames. The treatment of three-letter file extensions after wildcards is really just a side-effect of the 8.3 filename search behavior.

Consider the examples they gave:

"*.xls" returns both "book.xls" and "book.xlsx"

This is because the filename for "book.xls" (both 8.3 and long filename, since the name naturally complies with 8.3) and the 8.3 filename for "book.xlsx" ("BOOK~1.XLS") matches a query of "*.xls".

"*.ai" returns "file.ai" but not "file.aif"

This is because "file.ai" naturally matches "*.ai" while "file.aif" doesn't. 8.3 search behavior doesn't come into play here at all, because both filenames are already 8.3-compliant. However, even if they weren't, the same would still hold true because any 8.3 filename for a file with an extension of ".ai" is still going to end in just ".AI".

The only reason it matters whether or not the file extension in your search is exactly three characters is because the 8.3 filenames are included in the search, and 8.3 filname extensions for objects with long filenames will always have just the first three characters after the last dot in the long filename. The key part missing from the documentation above is that the "first three characters" matching is done only against the 8.3 filename.

So, let's look at the anomalies you're asking about here. (If you want any other strange behaviors explained, beyond your results for *.t and *.*.*.txt, please post them as separate questions.)

---

TL;DR:

Output of a search for *t includes 11.11.2007.txtGif and 2.txtGif.

This is because the 8.3 filenames match a pattern of *t.

11.11.2007.txtGif = 111120~1.TXT 2.txtGIF = 2BEFD~1.TXT

(Both 8.3 filenames end in "T".)

Output of a search for *.*.*.txt does not include 11.11.2007.txtGif.

This is because neither the long filename, nor the 8.3 filename, match a pattern of *.*.*.txt.

11.11.2007.txtGif = 111120~1.TXT

(The long filename doesn't match because it doesn't end in ".txt", and the 8.3 filename doesn't match because it only has one dot.)

Answer 3

This is the way that the Windows API works - you will see the same results if you use the dir command in a command prompt. This does NOT use regular expressions! It's pretty obscure...

If you want to do your own filtering, you can do it like so:

var filesEndingInT = Directory.EnumerateFiles(@"c:\test\").Where(f => f.EndsWith("t"));

If you want to use regular expressions to match, you can do it thusly:

Regex regex = new Regex(".*t$");
var matches = Directory.EnumerateFiles(@"c:\test\").Where(f => regex.IsMatch(f));

I suspect that you will want to let the user type in a simplified form of pattern and turn it into a regular expression, e.g.

"*.t" -> ".*t$"

The regular expression to find all filenames ending in t is ".*t$":

.*t$

Debuggex Demo