CGI Programming

1
CGI Programming

• Making web works at UCSC

2
CGI Common Gateway Interface

• CGI - oldest method for executing a program over
  the web. Very portable.
• A round trip between browser and server is
  required to update a page with pure CGI.
• Can add Javascript for increased interactivity at
  the expense of portability and complexity.

3
Anatomy of a CGI Script

• Gets called by Web Server when user clicks submit
  or follows a cgi link.
• Input is in environment variables and sometimes
  also stdin.
• Output is to stdout, primarily in html format.
• In the middle often access a database, do some
  calculations, etc.
• Sometimes in the middle create an image.

4
Challenges of CGI

• Each click launches program anew.
• User state has to be maintained by some special
  mechanism. We use the cart.
• Run from Web Server, harder to debug
• A program crash will cause an error 500 message,
  with truncated, often completely invisible
  output.
• Needs to run quickly enough to preserve sense of
  interactivity, ideally in less than 2 seconds.

5
CGI Input

• Input is contents filled out web form in one of
  three methods
• Get varval pairs in URL itself
• http//google.com/search?rlsen-usqgenomebrowse
  rbtnGSearch
• Post varval pairs in stdin
• Multipart stdin contains uploaded files as well
  as varval pairs.
• Routines in inc/cheapcgi.h transparently put data
  from all methods into a hash keyed by variable
  name.
• Typically we use a higher level interface,
  inc/hg/cart.h.

6
Persistence of State

• If have more than one web page on a site, want to
  preserve information entered in one page for
  later.
• Two basic methods exist for this
• Hidden variables - variable in form without
  associated controls.
• Cookies - Sent in http header. can be short or
  long term.
• Both hidden vars and cookies have limited space,
  so usually put most of data in a database, and
  just put key in var/cookies.
• Our database is called the cart.

7
How the cart works

• Two tables in database hgCommon - sessionDb and
  userDb.
• sessionDbs key is hidden var hgsid
• userDbs key is permanent cookie hguid
• Table contains key, blob with var/val pairs and
  date and use counts.
• Cart itself is a hash loaded in this order
  userDb, sessionDb, CGI variables.
• Routines in cart.h handle loading and access to
  this hash.
• You can explicitly add and remove things from
  cart as well.
• At end of program, cart is written back to tables.

8
Output of CGI Scripts

• Start out with HTTP header. Try
• htmlCheck getHeader http//genome.ucsc.edu/cgi-bin
  /hgTracks
• (htmlCheck very useful in general, based on
  htmlPage.c HTML parsing library.)
• Next comes HTML, most often a FORM.
• Often can include an image. Typically generate
  the image on the fly and include the URL for the
  image in the HTML.

9
CGI Script Output

• Typically HTTP header, and beginning/end of HTML
  is generated by routines in lib/htmlShell.c
  (vanilla) or hg/lib/web.c (UCSC look and feel)
• A wide variety of options are available, the
  newProg -cgi program will generate something
  basic.
• Use routines in cheapcgi.c to generate form
  elements such as drop-downs.
• Images are embedded with IMG tags or with INPUT
  image tags.

10
Images in CGIs

• Generate a file name in ../trash dir with
  makeTempName()
• Use vgOpenGif()/vgCloseGif() to start/end image.
• Draw with vgLine, vgBox, vgText, etc.
• Limited to 256 colors.
• Embed URL including trash dir in HTML.
• Theoretically could call a CGI that write image
  to stdout instead via URL. GMOD works this way.
• Use image map tags to make hyperlinks over image.
• Can also instead get mouse x/y position of image
  click.

11
Examples

• CGI that draws a rectangle, box, or line in
  user-defined colors.
• CGI that looks up common name from scientific
  name and vice versa.
• Use uniProt database and routines in
  hg/inc/spDb.h including spBinomialToTaxon.
• CGI that lets you paste in two sequences and
  returns alignment
• Use src/utils/faAlign and src/hg/mouseStuff/axtPre
  tty as starting point.