Kernighan and Pike challenge: how to put a slash in a filename

filenameshistorical-unixslash

I've just encounter the following question in Unix Programming Environment, the Kernighan and Pike's classic book on Unix (I found the below text on p. 79 of year 1984 edition, ISBN:0-13-937699-2):

Exercise 3-6. (Trick question) How do you get a / into a filename (i.e., a / that doesn't separate components of the path?

I'be been working with Linux for years, both as end-user and programmer, but I cannot answer this question. There is no way to put slashes in filenames, it's absolutely forbidden by the kernel. You can patch your filesystem via block device access, or use similarly-looking characters from the Unicode, but those aren't solutions.

I understand that Linux ≠ Unix, but the same principle should apply, since the system has to be able to unambiguously extract directory hierarchy from paths.

Does somebody know, what exactly Kernighan and Pike thought about when asking this questions? What was the supposed answer? What exactly is the 'trick'? Or maybe original Unix system simply allowed to escape this slash somehow?

UPD:

I contacted Brian Kernighan about the question and that's what he replied:

The answer is (or was) “You can't.”

Hence, Timothy Martin was right and gets the green tick.

Best Answer

Perhaps the answer is the same as part of the answer in this trick question:
How do you get down off an elephant? You don't. You get it from a goose.

From "The Practice of Programming" by Brian W. Kernighan and Rob Pike, Ch. 6, pg. 158:

When Steve Bourne was writing his Unix shell (which came to be known as the Bourne shell), he made a directory of 254 files with one-character names, one for each byte value except '\0' and slash, the two characters that cannot appear in Unix file names.

Related Solutions

Files – Understanding Filename and Pathname of a File

In POSIX terminology, a filename is the name of a directory entry. It consists of a non-empty sequence of bytes other than / or null. The term “pathname component” is synonymous with “filename”. A pathname is a string that can contain any non-null byte and that designates a way to locate a file. A pathname consists of a series of filenames, in which all but the last refer to a directory. Pathname resolution is the process of locating a file from a pathname.

For example, /home/tim/tim.pdf is a pathname. The last component of that pathname tim.pdf is a filename; it is the name of an entry in the directory whose pathname is /home/tim. The filename tim is itself the name of an entry in the directory whose pathname is /home; this file happens to be a directory. tim.pdf is also a pathname: any filename is a pathname that happens to have a single component and designates the file with that name in the current directory.

/ is a pathname that refers to the root directory. . is a filename that exists in every directory and refers to that directory itself. . is a pathname which happens to contain a single component and refers to the current directory.

You can think of a filename as a pointer within a directory to an inode that is a file in that directory. A pathname is a specification of where to find an inode. Each component of a pathname is a filename that points to an inode in the directory reached so far (if the file exists). Pathnames either start at the root directory (if they are absolute pathnames, beginning with /) or at the current directory (if they are relative pathnames, not beginning with /).

Note that in many texts, the word “filename” (or “file name”) is used to mean what POSIX calls a pathname.

How to put date and time in a file name

If you want to use the current datetime as a filename, you can use date and command substitution.

 $ md5sum /etc/mtab > "$(date +"%Y_%m_%d_%I_%M_%p").log"

This results in the file 2016_04_25_10_30_AM.log (although, with the current datetime) being created with the md5 hash of /etc/mtab as its contents.

Please note that filenames containing 12-hour format timestamps will probably not sort by name the way you want them to sort. You can avoid this issue by using 24-hour format timestamps instead.

If you don't have a requirement to use that specific date format, you might consider using an ISO 8601 compliant datetime format. Some examples of how to generate valid ISO 8601 datetime representations include:

 $ date +"%FT%T"
 2016-04-25T10:30:00

 $ date +"%FT%H%M%S"
 2016-04-25T103000

 $ date +"%FT%H%M"
 2016-04-25T1030

 $ date +"%Y%m%dT%H%M"
 20160425T1030

If you want "safer" filenames (e.g., for compatibility with Windows), you can omit the colons from the time portion.

Please keep in mind that the above examples all assume local system time. If you need a time representation that is consistent across time zones, you should specify a time zone offset or UTC. You can get an ISO 8601 compliant time zone offset by using "%z" in the format portion of your date call like this:

 $ date +"%FT%H%M%z"
 2016-04-25T1030-0400

You can get UTC time in your date call by specifying the -u flag and adding "Z" to the end of the datetime string to indicate that the time is UTC like this:

 $ date -u +"%FT%H%MZ"
 2016-04-25T1430Z

Best Answer

Related Solutions

Files – Understanding Filename and Pathname of a File

How to put date and time in a file name

Related Question