Follow Techotopia on Twitter

On-line Guides
All Guides
eBook Store
iOS / Android
Linux for Beginners
Office Productivity
Linux Installation
Linux Security
Linux Utilities
Linux Virtualization
Linux Kernel
System/Network Admin
Programming
Scripting Languages
Development Tools
Web Development
GUI Toolkits/Desktop
Databases
Mail Systems
openSolaris
Eclipse Documentation
Techotopia.com
Virtuatopia.com

How To Guides
Virtualization
General System Admin
Linux Security
Linux Filesystems
Web Servers
Graphics & Desktop
PC Hardware
Windows
Problem Solutions
Privacy Policy

  




 

 

The Art of Unix Programming
Prev Home Next


Unix Programming - A Tale of Five Editors

Now we're going to use five different Unix editors as case studies. It will be helpful to bear in mind a set of benchmark tasks as we examine these designs:

  • Plain-text editing. Manipulating plain ASCII (or, in this internationalized age, perhaps Unicode) files with no structure known to the editor above byte level, or perhaps line level.

  • Rich-text editing. Editing of text with attributes; these might include font changes, color, or other sorts of properties of text spans (such as being a hyperlink). Editors that can do this have to be able to translate between some presentation of the attributes in the user interface and some on-disk representation of the data (such as HTML, XML, or other rich-text formats.)

  • Syntax awareness. An editor that is syntax-aware knows that input events have a grammar, and does things like automatically changing the indent level when it recognizes the beginning or end of a block scope in a programming language. Editors that are syntax-aware also commonly highlight syntax with colors or distinguished fonts.

  • Output parsing of batch command output. The commonest case of this in the Unix world is running a C compilation from inside the editor, trapping the error messages, and then being able to step through the error locations without leaving the editor.

  • Interaction with helper subprocesses that persist and maintain state between editor commands. This capability, when present, has powerful consequences:

    • It's possible to drive a version-control system from inside the editor, performing file checkins and checkouts without dropping out to a shell window or separate utility.

    • It's possible to front-end a symbolic debugger inside the editor, such that (for example) when the run stops on a breakpoint the appropriate file and line is automatically visited.

    • It's possible to edit remote files within the editor, by having it recognize when a filename refers to another host (recognizing some syntax like /[email protected]:/path/to-file). Provided you have the right access, such an editor can automatically run a utility like scp(1) or ftp(1) to fetch a local copy, then automatically copy the edited version back to the remote location at file-save time.

All our case studies can edit plain text. (The reader should not take this capability for granted — there are many things called editors, such as ‘word processors’ that are too specialized to do this!) We begin seeing variable degrees of optional complexity in how they handle the more complex tasks.

ed(1) is the truly Unix-minimalist way of plain-text editing. It dates from the days of teletypes.[116] It has a simple, austere CLI, and there is no screen display. In the following listing, computer output is emphasized .

ed sample.txt

sample.txt: No such file or directory

# This is a comment line, not a command.
# The message above warns that the sample.txt file is newly created.
a

the quick brown fox
jumped over the lazy dog

.
# That was an append command, which added text to the file.  
# The dot on a line by itself terminated the append.
1s/f[a-z]x/dragon/
# On line 1, replace the first substring matching an f followed by a
# lowercase alphabetic followed by x with ‘dragon’.  The
# substitute command accepts basic regular expressions.
1,$p
the quick brown dragon
jumped over the lazy dog
# Print all lines from 1 to the last.
w

51

# That wrote the file to disk. The ‘q’ command ends the
# editing session.
q

Unbelievable as it may seem to a modern reader, most of Unix's original code was written with this editor. The reader with DOS experience may recognize here the original on which EDLIN was (crudely) modeled.

If one defines the job of an editor simply as enabling the user to create and modify plain text files, ed(1) is entirely sufficient for the job. Importantly to the Unix view of design correctness, it does nothing else. Many old-school Unix programmers half-seriously maintain that all editors with more features than ed has are simply bloated — and a few still who seriously believe this.

Appropriately, ed was Ken Thompson's deliberate simplification of the earlier qed[RitchieQED] editor — which was very similar (and the first editor to use regular expressions in the characteristic Unix way) but had multiple-buffer capability that Ken deliberately discarded. He judged it not worth the additional complexity.

A notable characteristic of ed(1) and all its descendants is the object-operation format of its commands (the session example shows an explicit range on the ‘p’ command). There is a relatively powerful syntax for specifying line ranges, either numerically, or by regular-expression pattern match, or by special shorthands for the current and last line. Most editor operations can be applied to any range. This is a good example of orthogonality.

Nowadays, ed(1) is primarily used as a program-driven editing tool in scripts — a role to which editors with more elaborate modes of interactivity are unsuited. There is a close variant called ex(1) which adds a few useful interactivity features such as command prompts; it is occasionally useful in rare cases when editing must be done over a slow serial line, or in certain unusual crash-recovery situations where the library support needed to run other editors is not accessible. For these reasons, every Unix includes an ed implementation and most include ex as well.

The sed(1) stream editor mentioned in Chapter9 is also closely related to ed; many of the basic commands are the same, though designed to be invoked through command-line switches rather than from standard input.

Almost all Unix programmers have strayed from the path of austerity and minimalist virtue enough to normally use editors that at least present a roguelike, screen-oriented interface. However, the fact that the religion of ed persists[117] says a great deal that is worth noting about the Unix mindset.


[an error occurred while processing this directive]
The Art of Unix Programming
Prev Home Next

 
 
  Published under free license. Design by Interspire