Objectives: By doing this problem set you will learn

• How to log into your development server
• Rudiments of Tcl
• How to run Tcl from the shell via tclsh
• How to write simple dynamic web pages using the ADP templating facility of AOLserver
• How to create, execute, test, and debug a .tcl page
• How to write a .tcl page that queries a foreign server
• Rudiments of SQL
• How to query Oracle from the shell using SQL*Plus
• How to write a .tcl page that queries Oracle
• How to personalize Web services by issuing and reading cookies
• How to read and write data in XML
• How to use SQL*Loader

This problem set requires you to learn a lot of new software, so make sure you get started early: plan to spend at least two or three sessions on it. There is nothing difficult here, but you need to work through the initial mechanics of using Tcl, SQL, and running the Web server; and this takes time.

Please feel free to use the on-line forum to ask your questions about this problem set or any other class-related problem. You can use the forum to view other people's questions, and provide or view answers. Also, we strongly recommend that you do as much of your beginning work as possible in the supervised labs, where the Lab Assistants on hand to help you.

Finding your Web server

Read Using the LCS Web/db Computing Facility and follow the instructions to log on to your Web server machine.

Getting Started with Tcl

Exercise 1: Running Tcl from the shell

Now define a recursive Fibonacci procedure in Tcl. Execute and test.

Hint: If you're writing Tcl programs of more than two or three lines, you'll find more convenient to type the code into a separate Emacs buffer (set the buffer to to tcl mode using m-x tcl-mode) and cut and paste from there into the Tcl shell buffer. Actually, you'll almost never be running Tcl from the shell in this course, but you will be writing lots of Tcl in Emacs as you create Web pages.

Exercise 2: Running Tcl from an (almost) HTML page

Augment the page so that (1) you add a $4000 South American Cichlid aquarium as an option, (2) you use a constructor procedure to build each aquarium element (instead of simply calling list), (3) you add an element to the aquarium for how many of each type of aquarium will be installed (4) you use procedures to extract type, cost and quantity from an aquarium element (instead of simply calling lindex), (5) you print out quantity-dependent subtotals and the grand total at the bottom.

Hint: If the code you write doesn't produce what you expect when you load the page (or, as is common, produces nothing at all), this is likely to be a Tcl error. To see if there was an error, look at the error log for your server, which is in the file /home/aol/log/yourvirtualserver-error.log.

Exercise 3: Simple Tcl pages

Notice how we use Tcl to read the form variables. Try out the form a couple of times, using your browser. Now debug the regular expression in simple-tcl-page-2.tcl so that it properly handles the names "Tammy Faye Baker" and "William H. Gates III".

Hint 1: It is easier if you don't try to do this in a single regexp. Use if then elseif then elseif ...

Hint 2: regexp has a side-effect. If you use a multi-clause if statement, make sure that you wrap your calls to regexp in braces so that they don't all get evaluated immediately.

Hint 3: Keep in mind that a match succeeds if the pattern matches a substring of the data. If you want to force your pattern to match the entire data item, you'll have to use an appropriate regexp to ensure this.

Exercise 4: Tcl pages that query foreign servers

We suggest querying barnesandnoble.com and www.1bookstreet.com. Your program should be robust to timeouts, errors at the foreign sites, and network problems. You can ensure this by wrapping a Tcl catch statement around your call to ns_httpget. Test your program with the following ISBNs: 0385494238, 0062514792, 0140260404, 0679762906.

Try adding more bookstores, but you may need to do kludges to make them work. For example, amazon.com and wordsworth.com tend to respond with a 302 redirect if the client doesn't give them a session ID in the query.

Extra credit: From which of the preceding books is the following quote taken?

"The obvious mathematical breakthrough would be development of an easy way to factor large prime numbers."

This would be a good time to take a break.

Getting started with SQL*Plus

Exercise 5: Talking to Oracle from the shell

create table my_courses (
	course_number	varchar(20)
);

Insert a few rows, e.g.,

insert into my_courses (course_number) values ('6.916');

commit;

select * from my_courses;

drop table my_courses;

Exercise 6: Tcl pages that talk to Oracle

To get AOLserver to load procedure definitions at server startup, you have to put .tcl files in your server's private Tcl library: /web/yourvirtualserver/tcl/. Create a file called "basics-defs.tcl" in this directory and define the following Tcl procedures:

• ad_header page_title -- returns html, head, title, and body tags, with argument enclosed within the title tags
• ad_footer -- returns a string that will close the body and html tags

Reload the quotations.tcl page and you get ... the same error! AOLserver doesn't know that you've added a file to the private library; this is only checked at server startup. Go to a Unix shell and "restart-aolserver yourservername" (this is the big hammer; it kills your server's Unix process so that Unix will restart AOLserver automatically). If restart-aolserver does not come back with "Killing 10234" or some other process ID, you'll know that you did not succeed (perhaps you made a typo when specifying your server name).

Reload the quotations.tcl page and you get ... a slightly different error! The program is trying to query a table that doesn't exist: quotations.

Go back to your sql shell and restart SQL*Plus. Copy the table definition from the comments at the top of the file quotations.tcl and feed this definition to Oracle. Go back to your Web browser and reload the page that previously gave you an error. Things should now work, although the quotations table is empty.

Use the form on the web page to manually add the following quotation, under an appropriate category of your choice: "640K ought to be enough for anybody" (Bill Gates). Note that it would be funnier if our table had a column for recording the date of the quotation (1981) but we purposely kept our data model as simple as possible.

Return to SQL*Plus and select * from the table to see that your quotation has been inserted into the table. The horrible formatting is an artifact of your having declared the quote column to be 4000 characters long.

In SQL*Plus, insert a quotation with some hand-coded SQL. To see the form of the SQL insert command you should use, examine the code on the page quotation_add.tcl. After creating this new table row, do select * again, and you should now see two rows.

Hint: Don't forget that SQL quotes strings using single quotes, not double quotes.

Now reload the quotations.tcl URL from your Web browser. If you don't see your new quotation here, that's because you didn't type COMMIT; in SQL*Plus. This is one of the big features of a relational database management system: simultaneously connected users are protected from seeing each other's unfinished transactions.

Preloading tables from data files

/web/yourvirtualserver/www/psets/basics/quotation-list.ctl

sqlldr userid=your-id control=quotation-list.ctl

Working with AOLserver and Oracle

Exercise 7: Improving the User Interface for data entry

Hint: In order to simplify debugging, you may find it useful to test your query using SQL*PLUS in the shell, before integrating the query into a Tcl page.

Exercise 8: Searching

Hint 1: Read about simple queries in SQL for Web nerds.

Hint 2: like '%foo%'

Hint 3: SQL's UPPER and LOWER functions

Personalizing Web services with cookies

• A user can "kill" a quotation and have it never show up again either from the top-level page or the search page.
• Killing a quotation is persistent and survives the quitting and restarting of a browser.
• Quotations killed by one user have no effect on what is seen by other users.
• Users can erase their personalizations and see the complete quotation database again by clicking on an "erase my personalization" link on the main page. This link should appear only if the user has personalized the quotation database.

You can personalize Web services with the aid of magic cookies. A cookie issued by the server directs the browser to store data in browser's computer. To issue a cookie, the server includes a line like

Set-Cookie:  cookie_name=value; path=/ ; expires=Fri, 01-Jan-2010 01:00:00 GMT

After the browser has accepted a server's cookie, it will include the cookie name and value as part of its HTTP requests whenever it asks that server for an applicable URL. Your Tcl programs can read this information using the AOLServer API

[ns_set get [ns_conn headers] Cookie]

You can see an example of how cookies are issued and read, by visiting the URL http://yourvirtualserver/psets/basics/tcl/set-cookies.tcl and examining the Tcl for file and the associated URLs check-cookies.tcl and expire-cookies.tcl. Observe how expire-cookies gets rid of cookies by reissuing them with an expiration date that has already past.

Reference: The magic cookie spec is available from http://home.netscape.com/newsref/std/cookie_spec.html.

Exercise 9

Hint 1: It is possible to build this system using an ID cookie for the browser and keeping the set of killed quotations in Oracle. However, if you're not going to allow users to log in and claim their profile, there really isn't much point in keeping data on the server. In fact, by keeping killed quotation IDs in your users' browser cookies, you've achieved the holy grail of academic database management system researchers: a distributed database!

Hint 2: It isn't strictly copacetic with the cookie spec, but you can have a cookie value containing spaces. Tcl stores a list of integers internally as those numbers separated by spaces. So the easiest and simplest way to store the killed quotations is as a space-separated list.

Hint 3: Don't filter the quotations in Tcl. It is generally a sign of incompetent programming when you query more data from Oracle than you're going to display to the end-user. SQL is a very powerful query language. You can use the NOT IN feature to exclude a list of quotations.

How about taking another break?

Sharing data with XML

In theory, people who wish to exchange structured data over the Web can cooperate using XML (eXtensible Markup Language), a 1998 standard from the Web Consortium (www.w3.org/XML/). XML has started to become widely hyped over the past year as "the next big thing on the Web", but in practice, hardly anybody uses XML yet (summer 2000). Fortunately for you in completing this problem set, you can cooperate with your fellow students: the overall goal is to make quotations in your database exportable in a structured format so that other students' applications can read them.

In order to cooperate, we need (1) an agreed-upon URL at everyone's server where the quotations database may be obtained; and (2) an agreed-upon format for the quotations. In point of fact, we could avoid the need for prior agreement by setting up infrastructures for service discovery and by employing techniques for self-describing data -- both of which we'll deal with later in the semester -- but we'll keep things simple for now.

We'll format our quotations using XML. XML structures consist of data strings enclosed in HTML-like tags of the form <foo> and </foo>, describing what kind of thing the data is supposed to be.

Here's an informal example, showing the structure we'll use for our quotations:

<quotations>
  <onequote>
    <quotation_id>1</quotation_id>
    <insertion_date>1999-02-04</insertion_date>
    <author_name>Bill Gates</author_name>
    <category>Computer Industry Punditry</category>
    <quote>640K ought to be enough for anybody.</quote>
  </onequote>
  <onequote>
  .. another row from the quotations table ...
  </onequote>
  ... some more rows
</quotations>

<quotation_id>
<insertion_date>
<author_name>
<category>
<quote>

Building a DTD

Our DTD will start with a definition of the quotations tag:

<!ELEMENT quotations (onequote)+>

<!ELEMENT onequote (quotation_id,insertion_date,author_name,category,quote)>

<!ELEMENT quotation_id (#PCDATA)>

Here's our complete DTD:

<!-- quotations.dtd -->
<!ELEMENT quotations (onequote)+>

<!ELEMENT onequote (quotation_id,insertion_date,author_name,category,quote)>

<!ELEMENT quotation_id (#PCDATA)>
<!ELEMENT insertion_date (#PCDATA)>
<!ELEMENT author_name (#PCDATA)>
<!ELEMENT category (#PCDATA)>
<!ELEMENT quote (#PCDATA)>

In this problem set, however, we won't use the DTD at all. You'll need to use only the informal XML quotations example.

Exercise 10: Generating XML

To get you started, we've provided the file example-xml.tcl. Requesting this URL with a Web browser should offer to let you to save the document to a local file, and you can then examine it with a text editor on your local machine. (Alternatively, your browser may already have some special behavior defined for MIME type application/xml. IE 5.5, for example, will automatically display the XML document.) The differences between our example and your program is that you'll need to produce a document containing the entire table and you'll need to generate it on the fly.

Exercise 11: Importing XML

• Grab /psets/basics/tcl/quotations-xml.tcl from another student's server using ns_httpget.
• Parse the resulting XML structure into records and the parse the records into fields.
• If a quote from the foreign server has identical author and content as a quote in your own database, ignore it; otherwise, insert it into your database with a new quotation_id. (You don't want keys from the foreign server conflicting with what is already in your database.)

Hints: You might want to set up a temporary table using create table quotations_temp as select * from quotations and then drop it after you're done debugging, so that you don't mess up your own quotations database.

Rather than having you work with DTDs and general XML parsers, we've gone for simplicity here by predefining for you a parser in Tcl that understands only this particular quotations XML structure. The procedure is parse_all. You have to install this file in your server's private Tcl library, /web/yourvirtualserver/tcl/, for this function to be callable by .tcl and .adp pages. (Don't forget to restart your Web server.) The parse_all procedure takes an XML quotation structure as argument and returns a Tcl list, showing the parts and subparts of the structure. To see an example of the format, use your browser to visit the page http://yourvirtualserver/psets/basics/tcl/xml-parse-test.tcl.

Note: these exercises are designed to familiarize you with XML. In most cases, XML processing should be done using Oracle's Java libraries. See http://www.arsdigita.com/doc/xml.html.

The Wide World of Oracle

Oraexercise 1: SQL*Loader

• Use Emacs to create a list of five records, each record containing a stock symbol, an integer number of shares owned, and a date acquired (in the form MM/DD/YYYY)
• Create a SQL table to hold these data:

  create table my_stocks (
         symbol	       varchar(20) not null,
         n_shares	       integer not null,
         date_acquired   date not null       
  );
  

• Use SQL*Loader to preload your data into the my_stocks table. You'll have to create an appropriate control file, as when you earlier preloaded the quotations table. You can probably guess the format of the control file by analogy with the quotations example, but you should at least be aware of the existence of the SQL*Loader documentation in the Oracle docs at http://philip.greenspun.com/sql/ref/utilities)

  Hint: If you leave blank lines at the end of the control file, the log file will contain error complaints about bad records with missing columns, since Oracle's implementation is too shoddily written to realize that a blank line isn't supposed to be a record definition. (What can you expect from a company with a mere $10 billion in annual revenues?)

Oraexercise 2: Copying data from one table to another

• Using only one SQL statement, create a table called stock_prices with three columns: symbol, quote_date, price. After this one statement, you should have created the table and filled it with one row per symbol in my_stocks. The date and price columns should be filled with the current date and a nominal price.
  Hint: Select symbol, sysdate as quote_date, 31.415 as price from my_stocks; .
• Create a new table:

  create table newly_acquired_stocks (
         symbol	       varchar(20) not null,
         n_shares	       integer not null,
         date_acquired   date not null       
  );
  

• Using a single insert into .. select ... statement (with a WHERE clause appropriate to your sample data), copy about half the rows from my_stocks into newly_acquired_stocks

Oraexercise 3: JOIN

Oraexercise 4: OUTER JOIN

Modify your statement to use an OUTER JOIN instead so that you'll get a complete report of all your stocks, but won't get price information if none is available.

Mini-Project

Choose a web page that provides some numerical data that is updated reasonably frequently, e.g., a couple of times a day. Two examples are:

• The temperature somewhere, extracted from some weather page.
• The sales rank of some particular book on amazon.com.

One of the things you might do is keep track of when the server you are accessing is up. For example, one of the interesting things about Amazon is that they often lose control of their server farm and database. (They write a lot of C code and one programmer's sloppiness can generate a catastrophic failure of the entire service.) You might want to build your system so that you can record (a) times when the server is unreachable, and (b) times when the page served some error message. For example, Amazon will send "Our store is closed temporarily for scheduled maintenance". (You'll sometimes get this during the middle of weekdays when they would definitely not have intentionally scheduled any maintenance.)

You can also display your data graphically as a chart. You can do this in a way that is reasonably browser independent, by using single-pixel GIFs and WIDTH and HEIGHT tags now. Grab the software in http://software.arsdigita.com/tcl/ad-graphing.tcl and put it in your server's private Tcl directory (/web/yourservername/tcl/). Read the docs at http://software.arsdigita.com/www/doc/graphing.html and then write code to generate a pretty chart of your data.

You might also provide a way to return your data as an XML structure, so that people can access it in building their own programs, without going through the same page-scraping shenanigans that you had to in collecting it. The idea that the Web could contain lots of data sources designed for programmatic access, so that people can combine them to build more elaborate programs, hints at the vision of Web services, about which we will have more to say anon.

Place your project on your web server on a page called project.tcl in your problem set directory, so that others can access it.

Turning in your work

Who Wrote This and When

This file is permanently housed at http://philip.greenspun.com/teaching/psets/basics/tcl/basics.adp.

This material is copyright 1999, 2000, by Philip Greenspun and Hal Abelson. It may be copied, reused, and modified, provided credit is given to the original authors with a hyperlink to this document.