Sindbad~EG File Manager

 12.18.06   Downloaded all source files produced by students in the CS240 class fall 2006 and began to try to make the project work
well enough to demonstrate the intended functionality in Leiden in January.   For that we don't need databases and student/teacher
logins.   

1.  There were two copies of constructor source.  Deleted the one not in the package.
2.  There was a copy of query.php as dbAccess.php with Sang Soo Kim's name as author (!).  Left it rather than change it to query.php.
3.  Deleted all reference to the Up button in constructor.java
4.  created theStack as specified
5.  Add the ActiveLine field to Script
6.  Got the In button to be enabled when it should be.
7.  Disabled database access by putting return into Query in dbAccess
8.  Deleted code in index.php that tried to access the database and made $include = 'start' to begin with
    so the login page is skipped.

12.19.06

9.  Discarded Constructor.java and started to rewrite it from scratch.  Completed "in" and most of "execute" in four or five hours.
10. Made isPrimitive static public in SymbolTable.java

12.20.06 

11.  finished "execute" and wrote displayScript. 
12.  Downloaded and installed Firefox 2.0 so the Error Console will be on the Tools menu.
13.  deleted some properties in stylesheets/index.css  that were causing error messages in the Error Console.
14.  In IE, but not in Firefox,  there is a "Java Console" under the Tool menu, that can display things written with System.out.println.
15.  Modified parser.java to accept an and-delimited list of variables after "return".  Constructor will retain all listed objects
     when execution returns to the caller; the first one will be the actual return value.  But execute1 doesn't do this, so 
     when we step over (as opposed to into) L = EquilateralTriangle(A,B,C),  we don't retain the triangle.   Moreover, in 
     CopySegmentScript, we don't even retain the first segment.  That's because when we decrease the level, we only 
     decrease it on the copy in theSymbolTable, which must be different from the one in theFigure.

12.21.06
16.  When a called script has a variable that clashes with a variable in the calling script (that is not passed as a parameter)
     then the value of that variable is not restored by Execute1  after it returns.  That is OK, according to the specs, but
     if that variable is an auxiliary return (i.e. mentioned in the "and" list after "return", this causes trouble.
17.  Implemented the "back" button successfully using new class State and a stack of States.  To do this I added the "copy"
     method in SymbolTable.
18.  Used the copy method also to make a copy of theSymbolTable to pass to execute1 when executing a
     script from Constructor.java.  This solves the problem in 16.  Now I can step through and back in the first two scripts.
19.  Well, not quite.  The triangle returned from EquilateralTriangle in the second script does not show up!  I spent two 
     hours trying to find out why.  Its visible field is false when its draw method is called, even though true shortly before that.
     That's because it's set to false in Diagrammer, line 236, because its argument D is already false.  But D is true just 
     before that.  Well, they must not be the same D, which also explains why D doesn't follow when A is dragged.   

12.22.06
17.   Added one line setIsInput(1) in Constructor.java after about two hours of debugging.  Without this, updateConstructions
      does not recompute D from its inputs, since it thinks D is an input point.  Now D does follow A when dragged.

18.   But the triangle in CopySegmentScript still doesn't show up, because its argument D is invisible so it is made invisible.
      How, where, and why is D made invisible?  There was a missing theFigure.add() in Constructor.java, but that wasn't the 
      problem.   Well,  D is only made visible after the return from the Triangle call,  but it's set as an arg of triangle T
      before that return.  If we just set all the args of multiple returns to be visible after the return, then triangle T
      shows up, but it does not drag correctly in CopySegmentScript, although it does in EquilateralTriangleScript.  Well,
      this behaviour proves that the D that occurs as an arg of T is NOT the same object as the D in theFigure.  Why not? 
      Because, in Execute1,  the multiple returns still have their args fields belonging to tempArgs.  So I put code in 
      Execute1 to rename the arguments of all the returns.   That does the job--now I can step through and back in both the 
      first two scripts.

19.   Now to make the inButton work.  The first test is to step in to EquilateralTriangle from CopySegment.  It's throwing a 
      ClassNotFound exception from Scripts.java looking for SegmentScript.class.  Well, there's no such file, of course, since 
      Segment is a primitive command.   I moved isPrimitiveCommand to Scripts.java, added primitiveReturnType, and made
      getReturnType work on a primitive or non-primitive command.

20.   Replacement of variable names in the called script doesn't preserve the  "P="  initial part of each line, resulting in 
      a null pointer exception.  The culprit is the toString method in the Command class.   Once I fixed that, and got the 
      construction of the Hashtable Subst correct in the in() method, it worked!  (at about midnight).

21.   But back() does not properly undo dragging and selection operations, and it seems that points can become 
      disconnected from the construction when back() is used.   That's because the clone() method of ArrayList does not 
      clone the objects.   I tried to fix this in the constructor of State.  But I still have the bug:  to see it 
      choose EquilateralTriangle, then execute, back, execute.  Now you can drag the points away from the segment!
      So evidently the args of the segment don't point to the same points that you're dragging.    I fixed this 
      by rewriting SymbolTable.copy() to fix up the arg pointers,   and using it in the constructor of State. 
       At 3 am the back button is working properly.

22.   Moving on to the third script, CopySegmentOntoLineScript,  we find that startData always produces points named
       A,B,C,D,  even when the script's parameters are P,Q,A,B.   

12.23.06
23.   Made startData a method in the Scripts class.  The default one will assign random locations, not closer together 
      than 40 pixels, to all the points,  and not do anything about objects of other types.   If a Script has
      parameters of other types than point,  or requires some relations between the points,  it should override
      startData  and call super.startData(),  and then fix up the data as desired and return it.

24.   Tested CopySegmentOntoLineScript and found crashes in SymbolTable.  The trouble was at line 531 of SymbolTable.java;
      the fix involved oldReturnLabel.   In addition, the script CopySegmentScript returns multiple values and so it must
      specify their return types.

25.   Next bug:  at the end of CopySegment,  if you drag by A,  point L doesn't change properly.  Before the return statement 
      is executed, everything drags fine.  The reason for this is that the arguments of L are now "dead", i.e. no longer
      in the symbol table.   After "return" is executed,  the construction and argument fields of the returned object need
      to be changed to the name of the  script to which the "return" statement belongs, and the arguments with which 
      that script is executed.   That will be the activeScript.   I fixed this in execute adding the part about actualArgs.
      Now, it works to drag by A at the end of CopySegment.  But, you hear the hard disk humming, probably due to garbage 
      collection?   

26.   Next bug:  Run script BisectAngle; step into EquilateralTriangle;  null pointer exception at the return statement.
      That exception was caused by actualArgs having the wrong length; a mere typo.   But then it still gives a null pointer
      exception.  At this point the original script variables have been renamed so you can't use getParameterNames to 
      look in theSymbolTable for ActualArgs.  Moreover, the  return statement doesn't seem to have been properly renamed.


12.24.06  
27.   Fixed this bug, rewriting the code for renaming variables in "in" to not rename the return variable.

12.25.06

28.   Still not working right; in CopySegment if you go "in" to EquilateralTriangle, "return" doesn't work right.
      That is because activeScript.getParameterNames is returning null for the parameter names!  I saw that bug 
      yesterday but somehow "it fixed itself".   
	  
1.10.07 

29.  Added fixupSubstForReturn in Constructor.java and called it.  This fixed incorrect behavior on 
return.   

30.  Added BisectAngle2Script.java, tinkered with menu wording in Diagrammer, etc.  

3.14.12   This year the site was moved to MonkeyBrains's server, and apparently it's now broken as a result.
Looking at index.php,  I see a test on $error,  when nothing has been assigned to $error.  I can't get any 
output from php on this page.  Apparently monkeybrains needs <?php   instead of <? .  
That would be because short_open_tag  is off in php.ini.   I changed that on the server in php.ini,
and then (when that had no effect) in php-ini-production.  That still had no effect.  Apparently I must 
also restart Apache, for which I used  

 /usr/local/etc/rc.d/apache22 restart
 
 That worked, enabling the short open tag,  but now the site is dead--nothing shows up. 
 The problem turned out to be, of course, Unix permissions.  In the meantime I learned that it is 
 OK to include a php file that contains HTML or other php code.  So eventually I got the toplevel screen to 
 show up;  but the applets aren't found.  So I changed CODEBASE in construction.php;  apparently the applets 
 were once one level up, and now they're under the main website directory instead.  So now it finds the 
 .class files ok but nothing is drawn.  The java files are  all in dynamicgeometry.zip. 
  Nothing is in the applets directory (presumably that's where the java files 
 should live).  
 
 3.15.12  It seems the difficulty is that Java isn't installed on the server.
 So I tried to do that, specifically to install openjdk6 following instructions found on the web. 
 I guess Rudy downloaded the files for me to start with, anyway they were there, but there were some hitches
 with files I had to to download locally, then upload to the server.   To get the Mac to not open zip files 
 automatically, use Safari Preferences | General  and uncheck "open safe files automatically".  Java has 
 to be built on the server using "make install clean".  There were zillions of scary warning messages from the C 
 compiler--it's a wonder this thing runs at all.  The makefile ran for maybe half an hour. 
 Finally it started echoing code instead of executing it,
 and I had to kill the build and try "make install"  again.   Eventually it completed, with a final message that 
 I should execute some commands that could not be executed, but javac works, and applets/AppletTest.html demonstrates
 that an applet runs on the server now.   Nevertheless MY applets do not run, they just hang up without initializing 
 properly.  The source code for these applets is under classes/org/dynamicgeometry/.   However,  I could not succeed
 to recompile them, either locally or on the server. 
 
 Preview in code does not work on my php pages, as it claims it can't find the included files.
 
 4.30.12   I modified AppletTest.html to use a "codebase" string and put a package command in the test
 applet's code (DrawingLines.java) and compiled it in org/dynamicgeometry/Test.  It runs fine.  Conclusion:
 the codebase command is not the problem.  
 
      I recompiled the applets in classes/org/dynamic/geometry/diagrammer and  /constructor  and /scripts.
All but one compiled successfully (that one script I renamed as ParallelogramToQuadrilateral.foo for now).
I checked that the class files now have today's date.  But the applets still don't load and run.

Maybe it's a browser problem. (Not likely as the test applet runs.)  Tried it in IE under Windows.  Lots of 
security warnings.  Told it to treat this site as trusted. Still get unknown publisher dialogs after which 
applet does not run.  

Maybe it's a Unix permissions problem:  Nope, I set all the permissions to 777 on the class files.  
I found out how to turn on the Java console in Safari; "class not found"  is what we get. 
Also in Firefox, which I downloaded.  Nevertheless AppletTest.html is working fine.  So what is 
different with AppletTest than with the  index.php?   Turned trace level to 5 in the Java console.
It's looking for the class files on a path including an incorrect segment ../  which is not needed
since the file is being 'included' by php, so when it runs, it thinks it's one level higher than it is.
It seems I didn't really manage to set all permissions to 777;  when that's done it DOES find 
the classes.  But now  Constructor and Diagrammer are throwing ClassNotFound exceptions.
Mysteriously, it works the same with 
  codebase="..//classes"   and codebase="classes".   It seems that exactly one of those should work.
Anyway, now it finds the classes.  But I get a null pointer exception when pressing the Execute button.
I recompiled Constructor.java and this problem disappeared.  But then, it came back; I'm still getting
"null pointer exception".

5.1.12  There were indeed duplicate copies of the code; and the codebase commands were misused.  I removed
all the codebase commands and made sure there's just one copy of the code.  Then I could see my debug messages
in the Java Console.  Soon I found the problem, and then
added code to execute() in constructor.java to make sure theSymbolTable is not null at startup.  Now the 
applet is working as well as it ever did.





  
     









 
 
 
 
 
 
 
 
 
 
 








    
     











To do (since 2007):   Change title from "Dynamic Geometry" to "Euclid's Constructions".
         Fix thumbnail display
         Fix flicker problem
         Fix need to double-click Execute the first time