Run statistics on typed shell commands

Programming Snapshot – Shell Stats

© Lead Image © Scott Rothstein, 123RF.com

© Lead Image © Scott Rothstein, 123RF.com

Article from Issue 243/2021
Author(s):

In the history log, the Bash shell records all commands typed by the user. Mike Schilli extracts this data with Go for a statistical analysis of his typing behavior.

How did the long command to connect to the database server work again? Shell power users know this memory problem and have taught themselves tricks over the years to repeat or resend modified versions of previously sent command lines. For example, the sequence !! repeats the last command (useful to prepend a sudo you might have forgotten); with Ctrl+R you can find and reuse commands that were sent further back in time by using search patterns. Bash remembers the entire typing history and shows it with the history command, which lists the tail end of what it has in memory (Figure 1).

Figure 1: The history command displays the last commands that were issued.

If you have ever taken a look at the .bash_history file in your home directory, you will know its little secret. This is where Bash simply appends every command sent, regardless of if it was successful or aborted with an error. If you ask for the last commands you entered, Bash simply looks there (Figure 2).

Figure 2: In the .bash_history file, Bash records which command was executed and when.

Timestamp It!

With a small change, Bash not only logs typed commands, but even adds the date and time when the commands were entered. To do this, you need to set the following environment variable (preferably in the .bash_profile init file):

export HISTTIMEFORMAT="%F %T: "

From then on, Bash adds a comment line with the Unix epoch stamp before each command recorded in the .bash_history logfile, while the history command prints the human readable date and time for each command in the display list (Figure 3).

Figure 3: If HISTTIMEFORMAT is set, the history returns the latest commands executed along with the timestamp.

This newly discovered data treasure trove begs for some kind of evaluation, of course. How about determining the days of the week on which a user was most diligent in terms of typing commands? Or how about identifying the most frequently typed commands in order to find new ways to reduce the future typing overhead by using abbreviations or shell scripts?

Callback as an Abstraction

Listing 1 [1] defines the histWalk() function, which finds the global history logfile for a user in the user's home directory, browses it line by line, and calls a user-provided callback function for each command entry. In this way, other analysis programs can define their own callbacks, pass it to histWalk(), and receive the history data including timestamps this way, without having to worry about the details of the history logfile's location or format.

Listing 1

histWalk.go

01 package main
02
03 import (
04   "bufio"
05   "os"
06   "os/user"
07   "path/filepath"
08   "strconv"
09 )
10
11 func histWalk(cb func(int64, string) error) error {
12   usr, err := user.Current()
13   if err != nil {
14     panic(err)
15   }
16   home := usr.HomeDir
17   histfile := filepath.Join(home, ".bash_history")
18
19   f, err := os.Open(histfile)
20   if err != nil {
21     panic(err)
22   }
23
24   scanner := bufio.NewScanner(f)
25   var timestamp int64
26   for scanner.Scan() {
27     line := scanner.Text()
28     if line[0] == '#' {
29       timestamp, err = strconv.ParseInt(line[1:], 10, 64)
30       if err != nil {
31         panic(err)
32       }
33     } else {
34       err := cb(timestamp, line)
35       if err != nil {
36         return err
37       }
38     }
39   }
40   return nil
41 }

The logfile opened for reading in line 19 by os.Open() is read line-by-line by the scanner created in line 24. To do so, it grabs the reader interface of the opened file, which is offered via the File structure, and takes in a new line with every call to Scan(); scanner.Text() then returns the text line as a string.

If the line starts with a comment character, it is a timestamp for a command in the subsequent log line. Accordingly, Listing 1 stores the seconds field in the timestamp variable (after cutting off the leading comment character) and sends the scanner off into the next round. If there is no hash sign at the beginning, it is a command line, and the program jumps to the else branch starting in line 34, where it calls the callback function entered by the user. As parameters, it passes in the previously saved timestamp and the currently read shell command.

First Class Functions

In Go, functions are first class citizens: They can be assigned to arbitrary variables or passed into other functions in parameter lists. For example, user programs can use the histWalk() function offered by Listing 1, assign it a callback function that is completely adapted to their special needs, and let it fill existing data structures in the local scope with the results.

The program logic should be easy peasy moving forward. Because of Go's strict typing, however, converting the previously read time value in seconds into an integer can try your patience, because the scanner reads it as a string. The ParseInt() function from the standard strconv package expects the string to be parsed as its parameter (line[1:] in line 29 cuts off the first letter – that is, the comment character), as well as the base of the integer number (10 for a decimal number) and the maximum number of bits (64). Back comes a int64 type value; this is an important detail: Otherwise the lights would go out on January 19, 2038, because by that date the supply of 32-bit integers for the seconds is exhausted.

One use for the histWalk() function is shown in Listing 2, which runs an analysis of the user's typing activities, broken down for the days of the week (Figure 4). Since the timestamp (stamp) passed to the callback is an integer, the standard time.Unix() function called in line 12 easily converts it to the Go internal time.Time format for time and date values. The Weekday() function called with the converted value then determines the day of the week for the given date. The int() wrapper converts the value that exists as a structure in Go's standard time package into an integer between   (Sunday) and 6 (Saturday).

Listing 2

dow.go

01 package main
02
03 import (
04   "fmt"
05   "time"
06 )
07
08 func main() {
09   var countByDoW [7]int
10
11   err := histWalk(func(stamp int64, line string) error {
12     dt := time.Unix(stamp, 0)
13     countByDoW[int(dt.Weekday())]++
14     return nil
15   })
16
17   if err != nil {
18     panic(err)
19   }
20
21   for dow := 0; dow < len(countByDoW); dow++ {
22     dowStr := time.Weekday(dow).String()
23     fmt.Printf("%s: %v\n", dowStr, countByDoW[dow])
24   }
25 }

Buy this article as PDF

Express-Checkout as PDF
Price $2.95
(incl. VAT)

Buy Linux Magazine

SINGLE ISSUES
 
SUBSCRIPTIONS
 
TABLET & SMARTPHONE APPS
Get it on Google Play

US / Canada

Get it on Google Play

UK / Australia

Related content

  • Astrology Hotline

    Because shell command sequences tend to reoccur, smart predictions can save you time typing. We first let the shell keep notes on what gets typed, before a Go program guesses the next command and runs it for you.

  • Effectively using Bash history

    You can do more with the Bash history command than just using the arrow keys. We show you how to use this command-line tool more efficiently.

  • Don't Know Much About History

    The versatile Bash history command can save you time and effort at the command line.

  • McFly

    When it comes to working at the command line, using Bash history effectively can save you time. McFly extends the Bash history's features and helps you find past commands more quickly.

  • File Inspector

    Spotify, the Internet music service, collects data about its users and their taste in music. Mike Schilli requested a copy of his files to investigate them with Go.

comments powered by Disqus
Subscribe to our Linux Newsletters
Find Linux and Open Source Jobs
Subscribe to our ADMIN Newsletters

Support Our Work

Linux Magazine content is made possible with support from readers like you. Please consider contributing when you’ve found an article to be beneficial.

Learn More

News