Bash: “history number” vs “command number”

bashcommand historycommand line

While Googling how to customize my shell prompt via the PS1 variable, I'm seeing tables of special characters that can be used. In particular:

          \!     the history number of this command
          \#     the command number of this command

"History number" seems to be more commonly used, and I know how to use commands like !523 to redo commands from history. But I can't figure out if "command number" has similar functionality. I've tried putting \# in my PS1 variable, and it seems to output the number of commands entered in a particular session (unlike \!, which persists after logout/exit).

Anyone know how to use "command number" in a convenient or meaningful way?

Best Answer

Bash's command number is for display only.

First, some background from bashref:

The command number and the history number are usually different: the history number of a command is its position in the history list, which may include commands restored from the history file (*note Bash History Facilities::), while the command number is the position in the sequence of commands executed during the current shell session.

Diving through the source, in parse.y we see that '\#' resolves to the global static variable current_command_number:

case '#':                                                                     
  n = current_command_number;                                                 
  /* If we have already incremented current_command_number (PS4,              
 ${var@P}), compensate */                                                     
  if (orig_string != ps0_prompt && orig_string != ps1_prompt && orig_string != ps2_prompt)
n--;                                                                          
  temp = itos (n);                                                            
  goto add_string;

which only has one other use: in eval.c, it's incremented upon running a command:

# ...
current_command_number++;                                                  

executing = 1;                                                             
stdin_redir = 0;                                                           

execute_command (current_command);

All that's kept is a number, not the actual command or even the equivalent history number. So, upon execution of each command, bash forgets what command associated with what command number, rendering the command number unusable for anything other than display and scroll reference.

Related Solutions

Bash – History command inside bash script

Bash disables history in noninteractive shells by default, but you can turn it on.

#!/bin/bash
HISTFILE=~/.bash_history
set -o history
history | tail …

But if you're trying to monitor activity on that server, the shell history is useless (it's trivial to run commands that don't show up in the history). See How can I log all process launches in Linux.

If you're debugging a script then shell history is not the best way to get useful information. A much better tool is the debug trace facility: put set -x near the top of the script. The trace is written to standard error.

Bash History – How to Set the Size of the History List in Bash to More Than 5000 Lines

This is the actual code that loads the history (from bashhist.c around line 260):

/* Load the history list from the history file. */
void

load_history ()
{
  char *hf;

  /* Truncate history file for interactive shells which desire it.
     Note that the history file is automatically truncated to the
     size of HISTSIZE if the user does not explicitly set the size
     differently. */
  set_if_not ("HISTSIZE", "500");
  sv_histsize ("HISTSIZE");

  set_if_not ("HISTFILESIZE", get_string_value ("HISTSIZE"));
  sv_histsize ("HISTFILESIZE");

  /* Read the history in HISTFILE into the history list. */
  hf = get_string_value ("HISTFILE");

  if (hf && *hf && file_exists (hf))
    {
      read_history (hf);
      using_history ();
      history_lines_in_file = where_history ();
    }
}

If the values of HISTSIZE and HISTFILESIZE are set, they will be used.

Readline, the library that actually handles input / line editing and history does offer facilities to put a cap on just how big the history buffer can grow. However, Bash does not place a hard ceiling on this where values any larger would be ignored, at least that I could find.

Edit

From comments, readline was indeed the culprit. I was looking (rather foolishly) at functional parameters:

there is a variable called history-size that can be read from the inputrc file. that variable sets the maximum number of history entries saved in the history list. I checked it's value in my local inputrc file to found it equal 5000. Setting it to a larger value solved the problem.

Best Answer

Related Solutions

Bash – History command inside bash script

Bash History – How to Set the Size of the History List in Bash to More Than 5000 Lines

Related Question