9.3.50: `system'

`system \system-command'

Tell the operating system to perform the indicated action. Whatever string follows the word `system' is passed directly to the operating system, after substitution of synonyms if any exist.

If your system command contains double-slashes, you must protect them from Gri (which will interpret them as comments) by enclosing in double-quotes, e.g. `system cat A | sed -e "s/foo//g" | cat > B'. (In the particular case of the `sed' command you could also do `system cat A | sed -e "s:foo::g" | cat > B'.

Note that `rpn' expressions are not evaluated, and variable values are not substituted before passing the string to the operating system. The exit status is stored in the builtin variable `..exit_status..'.

There are two ways to use the system:

• Assign output to synonym: The form `\synonym = system ...' does the system command and then inserts the output from that command into the indicated synonym.)

• Just run a command: The command `system ls' will list the files in the current directory.

For long commands, there are two approaches, the second preferred:

• Use continuation lines: String a lot of information onto one effective system line, using the `\' line-continuation character at the ends of lines. The problem is that it is very easy to lose one of these backslashes. The next method is better.
• Here-is syntax The here-is syntax of many unix shells is also provided. If the system command contains the characters `<<' followed by a word (with no space between!) then Gri will issue a system command which includes not only this line but also all succeeding lines, until a line is found which matches the indicated word precisely (with no leading space allowed). The `<< "WORD"' syntax is also supported, meaning that the operating system is being told not to mess with the dollar-signs -- needed in perl.

  Note: Be careful using this inside a new-command. Gri Recognizes the end of the body of a new-command by a line with `}' in the first column, and no non-white characters thereafter. If you system command happens to use a line with a curly brace (as in a loop in perl, for example), you must put whitespace before the brace. This won't affect the system command, but it will let Gri correctly realize that this is not the end of the new-command. For more information on new-commands (see Parsing).

  Caution: Before sending the string to the system, Gri first translates any synonyms present. Be careful with this, since system commands calling awk, etc, very often use backslashes for the newline character `\n' within strings. If you have a synonym whose name starts with `\n', you can get a conflict. For example, the awk command `print "foo\nbar";' should print a line with `foo' on it, followed by a line with `bar' on it, but it will instead print a single line with `fooMISTAKE', if you had previously defined a synonym as `\nbar = "MISTAKE"'. One way to avoid this mistake is to make sure any `\n' characters appear at the end of strings, and then simply avoid having a synonym named `\n'.

  Here is a Perl example.

  \m = "Foo bar" system perl <<"EOF" $a = 100; print "foo bar is \m, and a is $a\n"; print "BETTER: foo bar is \m, and a is $a\n"; print "Q: was a 100?\n"; EOF

Some more examples:

• To get the first 15 lines of a file called `foo.dat' inserted into another file called `bar.dat', you might do the following. Only the first method works; the second will fail because `.n.' will not be translated before passing to the operating system.

  \num = "-15" system head \num foo.dat > bar.dat # Following will not work correctly because .num. # will not be translated .num. = -15 system head .num. foo.dat > bar.dat

• Issue a unix command to get a listing of files in the current working directory, and pipe them into the `more' system command.

  system ls -l *c | more

• Store the date and time into a synonym, and use it in a title:

  \time = system date ... draw title "Plotted at time \time"

• Use `awk' to prepare a two-column table of x, ranging from 0 to 1 in steps of 0.1, and sin(x). The table is stored in a file whose suffix is the process ID of the Gri job. This file is then opened, and the data plotted. Finally, a system command is issued to remove the temporary file.

  system awk 'BEGIN { \ for (x=0; x<1; x+=0.1) { \ printf("%f %f\n", x, sin(x)) \ } \ }' > tmp.\.pid. open tmp.\.pid. read columns x y close system rm tmp.\.pid. draw curve

  Note: in unix, this command calls the Bourne shell, not the C-shell that is often used interactively. For many simple uses, the only difference from normal interactive use will be that `~' is not expanded to the home directory. For example, you should write

  system awk -f $HOME/foo/bar/cmd.gawk

  instead of the

  system awk -f ~/foo/bar/cmd.gawk

  that you might expect from interactive C-shell use.

  RETURN VALUE: Sets `\.return_value' to system status `N status'