Format of filenames passed to user commands

[diz]

Long-time x2 user running 3.2.0.2 x64 Unicode on Win7-64.

In a user command, I want to pass selected files/directories to a Java program. I am familiar with user commands, but this one is problematic.

the command is
>cmd /k java [JAVA_CMD_OPTIONS_HERE] $A

When a file or directory contains certain non-ASCII (Unicode) characters, the Java program is passed an altered filename.

E.g., if a filename contains the character ： [ U+FF1A Fullwidth colon], which is a legal Windows file character, the Java program is passed the filename with the fullwidth colon replaced by a regular colon, which of course isn't a legal Windows filename character.

With other legal unicode characters, the character is substituted or omitted entirely.

If I run the same program from the CMD prompt, e.g.,
java [JAVA_CMD_OPTIONS_HERE] m*

where the wildcard m* includes files containing unicode characters, the Java program is passed the correct filenames.

When filenames are passed to user commands via $F, $A, etc., are unicode characters altered in any way?

[Kilmatead]

When filenames are passed to user commands via $F, $A, etc., are unicode characters altered in any way?

Not as far as I have experienced. I cannot speak for the unicode-integrity of the java pipeline itself, but a quick test with a user-command of:

> cmd /k notepad $A

and passing a file with a specifically embedded wide-colon character in its path works fine for me (Win7 x64). Probably not related to this issue, but defining a user-command as "> cmd <prog>" is somewhat redundant as "$ <prog>" will execute via cmd directly.

Try passing the same files (via user-commands) through another programme (notepad, et al), and see if their path-integrity is preserved (as mine all are), which would, if nothing else, rule out x2 as suspect.

For clarity: Also note that x2 uses the standard Windows UTF-16 (2 byte) encoding, whereas (I believe) Java defaults to UTF-8 (2 or more bytes [anything above U+007F]), so if something in the java chain is re-interpreting the text, that may be pertinent. But that's just a guess.

[diz]

defining a user-command as "> cmd <prog>" is somewhat redundant as "$ <prog>" will execute via cmd directly.

cmd /k is necessary in this case: I want to examine the Alternate Data Streams of selected files and keep the result window open for inspection.

Perhaps a simpler example: say I want to easily mark a file as read-only. I create the following user command

Code: Select all

>attrib +R $F

In a test directory I create the following empty text files:
simpleTestFile.txt
：＊？“” ‘’.txt
The Naked Gun 33⅓ Final Insult (1994) 720p x264.txt

All filenames contain legal Windows filename characters. Run the attrib +R user command on each. In my case, it worked only for the first, despite the fact that the others are composed of legal Windows filename characters.

[Kilmatead]

You're confusing the user-command tokens. $F will supply only a single filename, regardless of how many are selected - also, it does not automatically cover spaces in the pathnames, so $F must always be surrounded by quotes ("$F") when used in a user command. (The multi-name $A token automatically does this for each selection, so must never be surrounded by quotes itself - but can't be used in this instance as attrib does not accept multiple consecutive filenames on a line.)

In other words, your command did not fail because of unicode:

If you want a user-command to iterate for multiple selections (one "cmd" each), use ">>" instead of ">".

In your example, considering some of those names contain spaces, were you to add quotes to the $F token and use it individually on each file, it will work fine, or just use: >> attrib +R "$F"

And yes, I tested this on your given examples, verbatim.

Off topic: "：＊？“” ‘’.txt" - fantastic filename.

[diz]

Good catch RE quoting $F.

Back to the drawing board RE the underlying cause of this problem. Will post any follow-ups as they arise.

[nikos]

cmd isn't Unicode and so funny characters may get altered if you use cmd. If you must use CMD, then instead of $A use $a that will pass the short 8.3 version of the filename

[Kilmatead]

Not true... CMD console is unicode-compatible out of the box these days (Win7+), though older installs (XP/Vista, or possibly US-EN installs) may still have trouble displaying the character set initially and require the traditional playing with codepage/font settings first. Of course inputting unicode is just as annoying as ever, but throughput processing is supported.

[diz]

Thank you all for the suggestions. Replacing "$F" with "$f" helped in another user command.

If I want to make all selected files/directories read-only, I can use

Code: Select all

>>attrib +R "$F"

But a window per file/directory briefly opens/closes in the background during execution. A bit unsightly. Using NirCmd's execmd option, the operation is executed without opening a cmd window in the background.

Code: Select all

>>C:\apps\x64\NirSoft\nircmd.exe execmd attrib +R "$F"

This works fine for nearly all files, but it chokes on some of the example txt files mentioned earlier. However, substituting "$f" allows it to process all files without a hitch.

Code: Select all

>>C:\apps\x64\NirSoft\nircmd.exe execmd attrib +R "$f"

On the subject of user commands: I often find myself navigating to Custom... User Commands... Organize... and finding that I want to 1) add a new user command, and/or 2) duplicate an existing command and modify it slightly for testing.

In the Custom Color Coding dialog one can modify an existing color coding scheme or add a new one. In User Commands... Organize it would be nice to likewise be able to add a new command without having to close the window and navigate to User Commands... Add new. It's a tiny annoyance, but those of us who work with user commands a lot would benefit from easier command management.

[nikos]

do you know that xplorer2 can set attributes too? press SHIFT+F12 on the selection

[diz]

do you know that xplorer2 can set attributes too? press SHIFT+F12 on the selection

Yes, I do, and it's a wonderful feature. But it still requires interaction with a dialog box (checking, clicking, etc.)

For operations like this I prefer to toggle an attribute on multiple files by selecting them and clicking a user command on the toolbar.

[Kilmatead]

For operations like this I prefer to toggle an attribute on multiple files by selecting them and clicking a user command on the toolbar.

Maybe I should add command-line attribute settings to this thing and I'd have more fans. Unfortunately for you, it's still just a dialog box, but could be extended easily enough to hide that.

Useful (for now) on the odd chance you would like to retain a detailed log of successful/failed/before/after attributes on up to thousands of individual objects. Some people really like logs of stuff on operations like that.

And unlike some people, I'm always open to user-suggestions.

[diz]

My Java Alternate Data Stream (ADS) viewer is working. In the unlikely event that other Java geeks are trying to pass files/directories with funky Unicode characters to Java from an x2 user command, as mentioned earlier, use, e.g., $a instead of $A to pass the paths as old-school 8.3 Windows filenames.

To resolve the 8.3 filename to the complete path, convert it as follows

Code: Select all

// sample: Java code to resolve full path from Win 8.3 path
public static void main(String[] args) {
  for (String s: args) {
    Path p83 = Paths.get(s); // path with Win 8.3 filename
    Path fullPath = p83.toRealPath(); // resolve 8.3 to complete path with Unicode chars, etc.
    if (Files.isDirectory(fullPath))
      // walk dir tree recursively
    else // process file
  }
}

[diz]

Just ran into a problem. I have a directory named "21 & 22 Jump Street 1080p", with an 8.3 format name of "21&22J~1".

The presence of the ampersand led to the Java program being passed a truncated filename.

The solution was to change the 8.3 format name, replacing the ampersand with an underscore, to "21_22J~1".

Should you ever need to change an existing 8.3 format filename, open a cmd window and type

Code: Select all

fsutil file setshortname <PathName> <shortname>