After returning from early-morning errand, created new MCRelat.java to
create that table if it didn't exist, and called it from the Setup
utility thereby finishing that, at 12:12.

Starting to implement logout function ... quick cross-circuit from
Loginout.logout to ActiveUsers.makeInactive done at 12:23, next to
call this from DispatcherServlet ... need nap at 12:32.

After nap and another errand, resuming work on logout function at
15:59 ...  I got the code written so it'd compile, but I didn't have
time to test it even once before uploading the files and then running
to catch the bus to class, which I missed, forcing me to take another
bus 15 minutes later, causing me to be 15 minutes late for the
class. (Did anybody notice I was late tonight May.31?)

After class, in the lab, I downloaded the files and tried the Setup
program: The windows system properties weren't set up yet, so I had to
copy them from the Setup output to the properties file. Then I
discovered that backslant in a properties file isn't treated as a
regular character, but instead combines with the next character, so I
needed to change all backslant to two copies of backslant. Then that
part of Setup worked.  The next problem was that the routine to create
tables in the database was reporting that the CREATE TABLE ... command
had failed, because the return code from my wrapper around
executUpdate was apparently catching a SQL exception and returning -1
to the caller. But in fact MicroSoft ACCESS was violating the API by
returning -1 instead of 0 in that case.  So I had to change my code to
notice the -1 and change it to 0. Then that part to the end of Setup
worked perfectly! Then I tried Top1 which is the NULL toplevel test
rig just for the starting case where it gives me a login form, and it
worked, including converting to swing. I didn't have with me the file
of test data, so next I went directly to the TomCat server. The login
form came up fine, but when I filled it out with correct info and
submitted it, the server wouldn't accept the connection, because
whereas the port number is 8000 on Linux, it's 8080 on Windows, so
whereas the URL that I manually composed by modifying the
sample-program URL was correct, the URLs generated by the program
itself for action upon submitting a form was still the Linux URL. So I
fixed the port number in the URLs in the program, but then it gave me
file not found, because I needed to fix the path to the servlet
also. After I did that, it worked, giving me the main screen showing
options (send message, show report, and log out). Unfortunately I got
servlet exceptions when I tested invalid login, and when I tested
logout. Then I checked the time and noticed I had only 10 minutes left
before I had to catch the bus home (or miss that bus and have to stay
more than an hour later for the next bus), so I didn't have time for
any more testing or even for e-mailing myself the changed files so
that I could merge the changes with the master files on my laptop.


It's now June.01 day-time. Just after 7 AM I woke up and since last
night at class the instructor started talking about EJBs and asked me
about my experience with them way back in early April, which was a
long time ago, nearly two months ago, I decided this morning to check
if the EJB application ConverterApp.ear was still around, yes it was,
and to look around in it to see what parts it has, which took more
than an hour before I just gave up because it takes 5-10 minutes for
the screen to be updated almost every time I click on a new tab within
deploytool. Finally after giving up looking around, I deployed the
application, which took about ten minutes. Then I tried the Web
client: It took a very long time to start up, but it worked again,
using the original code before I changed the exchange rates in the
source and compiled version. Then I tried the application client. The
names of two files on the disk didn't start with the same case as the
corresponding names in the command I copied and pasted, so I renamed
the two files to match the command (actually one file specifically
mentionned in the command, and the other just so all files would have
the same-case first letter):

$ mv converterApp.ear ConverterApp.ear
$ mv converterAppClient.jar ConverterAppClient.jar
$ ls
Converter.java	  ConverterAppClient.jar  ConverterClient.java	index.jsp
ConverterApp.ear  ConverterBean.java	  ConverterHome.java	index.txt

Then when I re-tried the command, this happened:

runclient -client ConverterApp.ear -name ConverterClient -textauth
Initiating login ...
Username = null
Enter Username:guest
Enter Password:guest123
Binding name:`java:comp/env/SimpleConverter`
Caught an unexpected exception!
javax.naming.NameNotFoundException
	<<no stack trace available>>
Unbinding name:`java:comp/env/SimpleConverter`

I have no idea how to find out what name was not found, I thought
maybe there was an internal reference from one of the files or naming
authority to that mis-capitalized file, so it needed to be different
from the other files, so I renamed it back:

$ mv ConverterAppClient.jar converterAppClient.jar

Then I retried the runclient command, but got exactly the same error
as before. I give up. Time now 09:09, appx. two hours burned on this
worthless exercise.

Back to work on Lab#3, specifically trying to remember the changes I
needed to make on campus last night to make the program slightly work
on the lab's computer. Put dummy fields in properties file until I can
get the real info on campus (put the os.name and java.version as best
I can remember, and put real junk for java.io.tmpdir) ... done at
09:13. Copied mainUrl from DispatcherServlet.java to lab3.properties
under name linux.mainurl, and made dummy entry for windows.mainurl as
best I can remember it. Made new method in Lab3Props.java called
getMainUrl which simply fetches that property (per correct environment
linux or windows) and returns it. Added a unit-test for that new
method in main. Compiled and ran test, works, at 9:23. Edited
DispatcherServlet.java to call that in the four places where
appropriate, but felt totally drowsy at 09:27 and had to take nap.

After nap and errands, now 14:16, got that working, then debugged the
logout feature and got that working, now implementing next use case of
presenting menu of cabs as checkboxes and allowing user to send
message, at 15:48 ... at 15:56 I absolutely needed another nap
immediately, finished at 17:28 and resuming work ... finished writing
code in Cabs.java to return vector of vector-pairs [cabID#, cabName]
and test rig showing it works at 17:42. Next to call that from servlet
to build sequence of checkboxes ... wrote routine in ClientWriter to
build a checkbox structure, and tested it, done at 18:59. Upgraded
ClientWriter.toHtml to handle case of checkboxes, done and tested and
working at 19:37. I'll skip conversion to swing for the moment.  Added
new feature to unwrapper (for NULL mode) where the HTML can be written
to a file on temporary directory, whereupon I can have a Web browser
looking at that file and simply re-load the page whenever I want to
see the latest version of HTML overwritten there. So using that new
feature and lynx to view it, I've verified the form listing all the
taxicabs as checkboxes and with a place to compose message text is
working, as of 22:05. (If I had thought of this idea a couple weeks
ago, I wouldn't have needed to implement conversion to swing at all as
a debugging tool.)

Next, at 22:24, to write code to take the result of filling out that
form and actually (pretend to) send messages and register them in the
database ... First I needed to implement multiple values for the same
tag, to handle multiple checkboxes all with the same name but with
different values. Most of the code is written but it's bedtime so I
won't be able to finish it until morning. Now 23:56.


It's now 2005.Jun.02 at 08:18, starting to write unit-test for the
code I wrote just before bedtime last night. Wrote function to print
out entire contents of HashMap in legible form.  Also included
decoding of MIME-coded-form-contents for each tag and value extracted
from tag=val&tag=val&... string. Tested and working at 09:31.

Next, starting to write code to use all that to process the submitted
send-message form ... completed code in MCRelat.java to add
relationship record, tested working at 10:46. Next, starting to write
new file DBUtil.java per specs in dj3dbutil.txt, starting with
sendMessage(...) ... got code working and tested as of 12:08 but need
to turn off verbose trace output ... done at 12:23. Next to call that
from the servlet ... done but bitten by apostrophe in the text of my
message which broke SQL update command. Tried putting \ before the '
but SQL doesn't recognize that convention. Finally I gave up and
re-worded to say "do not" instead of "don't" and it worked.  It's now
12:58. Need some code cleanup ... done at 13:28.

Now working on showing report of last N messages ... got business
logic working in DBUtil as of 14:49, now to reduce trace output and
call from servlet ... done at 15:21. Next to put in the button to take
user to menu screen again ... all working (in NULL test rig) as of
15:44.

Next to startup J2EE server and do a live test ... copied files,
server up at 16:03. Ready to try live test ... gave up at 16:09 when
server still hadn't put up first Web page, already had no nap and need
shower and late for class and still need to upload files ...  so what
I did was to initiate stoppage of J2EE server and take shower while it
was stopping, then initiate stoppage of cloudscape and dry while it
was stopping, then rush to collect and zip and uuencode and upload the
latest files, not bothering to uudecode or e-mail them yet, then
quickly log out and shut down Linux while getting dressed to catch the
bus, to arrive only 2 minutes after the start of class but the
instructor had already started talking.

After class, in the lab, I TELNETted to my shell account to uudecode
what I had uploaded then e-mail to my Yahoo! Mail account, then log
into my Yahoo! Mail account to download the zip and then unzip it,
then reconciled the original and modified versions of Tuesday's files
to see what changes had been needed to make the earlier version work
on Windows, then folded those changes into the new files from Linux.
Before testing anything, I e-mailed from my Yahoo! Mail account
attachments of all the files that had changed, so that later I would
be able to reconcile them on Linux and finally have a standard version
of the files that would be the same on both systems. Then I
established the MicroSoft ACCESS data source, ran Setup, then ran the
NULL test loop briefly to see if things worked at all, they did.  The
started TomCat server and tried the servlet: It worked!!


Today, Friday Jun.03, I finally tried downloading BeanShell via Unix
and dialup modem to my laptop, and tried running it here in
special-window mode: It worked, although copy&paste with emacs or X
window don't work, although copy&paste with gNotePad and gEdit does
work, so if I need to copy text between the two environments I can
save text to a file from one envrironment and then load that file with
an editor in the other environment and copy from there, a big pain for
snippets of code but tolerable if I need to copy a large chunk of code
across. Also if I run a program that does terminal I/O, it's all
confused because BeanShell is doing I/O in the special window while
the program it's running is doing I/O in the X window from which I
started BeanShell, and it's difficult to know where to type anything
and how to abort anything.

Then I tried running BeanShell terminal-mode on my shell account on
Unix ISP: It bombs immediately at start-up:
Exception in thread "main" bsh.InterpreterError:
 internal error 1 in This: java.lang.ClassNotFoundException: bsh.JThis
  at bsh.This.getThis(Unknown Source)

Then I tried running BeanShell terminal-mode on my laptop: It works!
And of course copy&paste between emace and terminal-mode in X terminal
works the same as always. As a real test, I copied individual lines of
code from a swing application from the previous Java class to emacs,
edited them there, then copied individual modified lines of code to
BeanShell terminal-mode, and as a result I'm sitting here with a
JFrame visible on-screen. So it seems I can use BeanShell for
LISP-REP-style development for real if I want. So I'm going to try
future bottom-up tool-building development of the current
Distributed-Java Lab#3 project using BeanShell to avoid having to wait
a full minute or so every time I want to test one or two new lines of
code, running a compile cycle only when really necessary. But first,
I'm totally exhausted from no-nap-yesterday, so I needed to take a
long nap mid-day.

It's now 17:14 as I start using BeanShell for continued work on Lab#3
taxicab messages. Starting work on use case #3 (display per-cab
totals) ... at 19:38 I have the business logic finished and debugged
in BeanShell for collecting all the data and returning a Vector of
3-Vectors each of form [CabID# CabName MsgsPerCab], and in the process
I extended Sexpr1.toSexpr to handle int[] just as it already handled
String[], and what a relief, never had to run javac once the whole
time! Now at 19:40 to run javac on all the changes files to see if we
(BeanShell and myself) missed anything javac would catch ... in Sexpr1
there were no compiler errors. In DBUtil there were a few minor errors
such as where I changed the name of a variable halfway through
development and was still referring to the old name from one place,
and where I set up a variable in one method but failed to set it up in
another method where it also was used. And then when I fixed those so
that javac could compile more deeply, it discovered that I had failed
to catch java.swl.SQLException every place I did something with a
ResultSet. Fixing those, at 20:01 no more compiler errors. No need for
more unit tests on that code until I write the one line of code in
DispatcherServlet which calls the toplevel method of the new business
logic and unit-test that one line of code as part of developing the
servlet code.

Starting work on on that servlet code ... aha, the first serious bug
in BeanShell: The following line of code crashes BeanShell all the way
out to terminate fatally out to the shell:

bsh % 	wrReq = HSA.wrapRequest(objReq);
Exception in thread "main" java.lang.NoClassDefFoundError: javax/servlet/ServletResponse
	at java.lang.Class.forName0(Native Method)
	at java.lang.Class.forName(Class.java:195)
	at bsh.classpath.ClassManagerImpl.classForName(Unknown Source)
	at bsh.NameSpace.classForName(Unknown Source)
	at bsh.NameSpace.getClassImpl(Unknown Source)
	at bsh.NameSpace.getClass(Unknown Source)
...

So when that happens, I lose my BeanShell entirely, and need to fix
the problem then re-start the BeanShell. I think the problem was that
it needs the j2ee.jar included in classpath. I'll do that and try
again at 20:34 ... yes, that fixed the problem. Continuing work at
20:41 ... OK, I've gotten to that one line of code that unit-tests all
that I was working on before, but now in javac compiled form:

bsh % v = DBUtil.getPercabNamesTotals();
<[[1, Johnny, 4], [2, Mikey, 7], [3, Janice, 8], [4, Monica, 1], [5, Judy, 8], [6, Murphy, 4]]>

and of course it works just as well now as it did when the pieces were
built by BeanShell. Continuing at 20:45 ... now I need to write code
to generate multi-UI structure for a table, but first I'll manually
build a very small table just to make sure lynx renders it correctly
built, checking in lynx now ... no good, lynx doesn't bother to line
up columns whatsoever:

   Name Relationship Location 
   Robert (self) Sunnyvale
   Heather Beloved (unknown)
   Elton Father Oregon

so I tried it in Netscape, and it worked correctly, but when I copy
from Netscape and paste here, I get this nicely-lined-up mess:

 Name
         Relationship
                     Location
 Robert
         (self)
                     Sunnyvale
 Heather
         Beloved
                     (unknown)
 Elton
         Father
                     Oregon

Anyway, now that I know that I know how to do tables correctly, and
that Netscape is rendering them correctly so I can see if my software
generates them correctly, I can proceed with the code, at 20:59 ...
at 21:46 I'm getting a strange exception I didn't get when I did the
same thing before: java.sql.SQLException: no such object in table This
is when I repeat the query: SELECT max(CabID) FROM Cabs which works
just fine when I re-do it manually in cloudscape -isql, but repeatedly
fails when I re-do it from my program code. Hmm, when I re-run the
Setup program, it claims that table Cabs doesn't exist, and when it
tries to create the table it gets that same exception as above.  I
tentatively conclude the connection to cloudscape is in a funny state,
or cloudscape itself is in a funny state. Let me try the setup program
directly from the shell ... it says everything is fine, all tables
exist and expected number of records in each, agreeing with cloudscape
-isql.  Next thing to try is to kill BeanShell and re-start it, but
first I tried to kill cloudscape -isql whereupon it gave me that same
strange error. Continuing to kill BeanShell ... done, restarted and
ran Setup with no problem, re-ran the method-call where it failed
before and now it works, re-defining the last method I did before it
died, re-running that method-call again where it failed and now it
works, finally running for the very first time the unit test of that
new method ... it works, giving the multi-UI structured object representing
a FORM:

("TABLE" ("ROW" <1> "Johnny" <4>)
	 ("ROW" <2> "Mikey" <7>)
         ("ROW" <3> "Janice" <8>)
	 ("ROW" <4> "Monica" <1>)
         ("ROW" <5> "Judy" <8>)
	 ("ROW" <6> "Murphy" <4>))

Next at 22:08 to write code to convert that to HTML but first a break
...  but too tired to do anything except use the last bit of my energy
to wash some dishes then go to bed. But can't sleep, so at 01:29
resuming work ... done unit-testing code for basic 3-column table at
02:39, now to write code to convert to HTML and also include column
headers in the original structure ... done in crude form at 03:12,
next to incorporate into servlet so NULL test rig will write to test
file ... at 03:17 I'm getting that "java.sql.SQLException: no such
object in table" again. Time to re-start BeanShell, and while I'm at
it I might as well re-compile sources which changed ... no compiler
diagnostics in ClientWriter, and in DispatcherWriter only one
method-call I prematurely activated before it was defined, and three
undeclared variables, done at 03:31. Time to go to bed again, maybe
for real this time.


Now 2005.Jun.04 at 09:55 as I resume work on whatever I was working on
in the middle of the night ... done putting pieces together so I could
call the business logic then the unwrapper to write the HTML to /tmp/
file, and viewed in Netscape, and table looks correct as of 10:31.
Next to finish putting the one toplevel call into the servlet file and
compiling all changed files ... done at 10:38 (no compiler
diagnostics), next to set up test data to invoke the new use case ...
done, ran NULL-mode test rig, wrote /tmp/ file, looked at it in
Netscape, looks fine!

Next I need to include the button at the bottom for getting back to
the main menu, and replace the column head for "Msg count" to be a
button that sorts by that column ... but first I cleaned up the code
by making sure there's a comment at the start of each method
explaining briefly what its purpose is and what actions it performs,
found method to build to-main-menu button, incorporating it now ...
done, compiling ... done, running NULL-mode test ... done, wrote /tmp/
file, looked at in Netscape, looks perfect at 11:12.

Before starting to implement column head to sort by #messages, I'm
going to shut down cloudscape and emacs, shut down and restart Linux,
restart cloudscape, and start J2EE server for first time since aborted
attempt yesterday afternoon, first real time since a couple days ago,
and run a *live* J2EE HttpServlet test, starting shutdowns at 11:15
... Linux was fully down at 11:18 and starting back up ... oops I
forgot to go into super-user mode before trying to start J2EE, so it
died when it couldn't start JMS, so I had to re-start that part ... at
11:28 J2EE was up and I started bringing CloudScape up ... at 11:30
that's up, running Setup ... AOK, started up lynx and clicked on link
to start servlet at 11:33, waiting for J2EE server to deliver the
initial page ... ops, I forgot to copy the changed class files from my
development directory to the system directory, doing that now just as
server is offering cookie to lynx at 11:37 ... done at 11:43, now to
try lynx ... accepted cookie at 11:45, filled out log-in form and
submitted it at 11:46, waiting for server to process the form ...  got
main menu at 11:49, clicked on send-message button at 11:50, waiting
for server to bring up send-message details form ... got form at
11:54, filled it out (send to Mikey and Monica) and submitted during
same minute, waiting for server to confirm message sent and show last
N messages ... got that at 12:00, clicked on main-menu button, got
main menu quickly at 12:01, clicked on button to show msgstotals per
cab at 12:02, waiting for response ... at 12:05 I got servlet
exception from not-yet-implemented exception in yesterday's code, so
it seems the server has cached the old code before today's changes, so
at 12:06 I initiated J2EE shutdown ... at 12:15 I got first response
from the two programs that they were shutting down service, at 12:17
shutdown finished and I invoked starting J2EE back up ... up at 12:22,
started lynx, clicked on button to run servlet at 12:23, waiting for
server to respond ... at 12:28 it offered cookie, I accepted cookie
and filled out login form and submitted at 12:29, waiting for server
response ... got already-logged-in warning and offer to attach to
already-existing login session and main menu at 12:31, clicked button
to show report of per-cab message totals (the new code of today),
waiting for server to respond ... got report (bad format because I'm
in lynx) at 12:37, clicked on button to return to main menu ... got
main menu at 12:41, clicked log-out button ... at 12:44 it gave
confirmation of logout and showed new log-in form, end of test.

Conclusion: after debugging in BeanShell, module compiler diagnostics
it worked perfect first time in NULL-mode test rig, and then with no
further changes but an hour and a half of time it worked perfectly in
live J2EE servlet-container session. Meta-conclusion: I wish I had
thought of trying BeanShell a couple months ago, before this huge lab
assignment started, so I could have used it the whole time, would have
cut development of NULL-mode test rig to one week instead of two, but
I'm glad I thought of it now. Now I'm going to shut down J2EE server
and take a nap before starting coding of use case for clicking on
column header to sort by contents of that column.


After nap and washing dishes etc., at 15:03, and I'll split this log
at this point and update the index file and upload both ...