Starting with Some Small Stuff

Administrivia

* I've had a request to move this to Tuesdays.
  * Staying on Mondays.

Review

* What did we learn last week?
  * Something good to know: Redirecting stderr to stdout
  * Automate for efficiency
    * If you're doing something once, by hand may work
    * If you're doing something multiple times, writing a helper
      amortizes
    * tr is a useful tool for doing character-based replacement
      or deletion
      * Generalize: You may not need to write programs; tools may
        already be there, so learn the tools
    * od is also a useful tool

Simple *nix Tasks, Continued

* Given a CSV in which each line has 
  LastName,FirstName,Assignment,NumericGrade 
  find the lines with the five highest grades on homework 2.
  * Observation: 
    * Need a way to identify the grade part
    * Need to know how the programmer identified hw2 ("hw2")
  * Strategy: Identify subtasks
    * extract only the lines that have hw2 (however it is represented)
      grep ",hw2," grades
      # Hidden assumption: No one has a name or numeric grade of "hw2"
    * sort lines by grade
      sort -t, -k4 -g -r
    * get the first five lines
      head -5
  * So, pipe them all together
  cat grades.txt | grep ",hw2," | sort -t, -k4 -g -r | head -1
    * Not only do we need to know tools, we need to be willing to
      combine them.
    * Getting close to write-only code.

* Given an HTML file, find the URLs of all images.
  * Assumptions: no more than one image per line, complete tag per line
  * Identify subtasks
    * Find the image tags in the file using grep
        cat file.html | grep -i "<img"
    * Identify the URLs on those lines (all of them?)
        src="..."
      and chop off everything outside of the quotation marks
        sed (stream editor) is useful for this
        sed -e 's/^.*src="//' | sed -e 's/".*//'
    * Generalize: What if there can be multiple image tags on one line?
      * Don't throw away the rest of the line; want to iterate over
        all of the image tags on the line.
      * Or ... make sure that each image tag is on a separate line
        tr '<' '\n'
        * Moral: Sometimes it's okay if you turn everything into a nail.

Important *nix Tools
* We'll make a list of *nix tools every programmer should know at least a little about.

* Your shell (bash, csh, sh, tcsh) and its built-in commands
  * cd, ls, mv, cp, pwd, cwd, pushd, popd, history, mkdir, rmdir, rm
  * ! stuff
* cut - rearrange columns in a file (e.g., CSV, TSV)
* fmt
* grep: Look for all the *lines* of a file that contain a particlar
  piece of text (lots of variants)
* head: grab the first few lines of a file
* make
* more and less, more or less - page through data
* od - look at data
* sed - stream editor (general purpose; until expert, use for a few
  small tasks, like substitution)
* sort
* tail: grab the last few lines of a file
* tr: translate (or delete) single characters

Make

* Why Make: 
  + If you need to type complex instructions, you should write them down.
  + If you've written them down, you should be able to run them automatically.
  + Generalize: And, hey, many instructions are common anyway.
* What else do you know about Make?
  * Make asks you to store commands in a file typically called Makefile
    (or makefile)
  * There are a lot of options of what you can put into a Makefile.
  * It is possible to write a 50-page Makefile
    * These days, people use Automake to write those long Makefiles.
  * What are these common things we put in a Makefile when we run it
    by hand.
    * Things you want to make and the instructions for doing them.
    * These include:
      clean:
        How to clean up after your compilation.
      default:
        What you want to do if you're not sure what the central part
          of the program is.
    * General commands
    * Options to general commands


Looking Ahead

* What do you want to do next week?  Some time this semester?
  * We will do the c preprocessor next week!
* We'll also make a list of C concepts that every programmer should
  know at least a little bit about.

Question: Is there a more efficient way to make a copy of a database
for testing?
* Traditionally, dump to a text file and then reload.