Code Generation for Assignment 5
New comments and grammar elements for this assignment are in green. Characters that
appear in the generated code are in red.
Pixels are conceptually a 3-tuple with the red, green, and blue color components. Internally,
they are stored in a packed form where the three values are packed into a single integer.
Because each color component must fit in 8 bits, the values are in [0, 255]. When the color
components are provided in the form of an ExpandedPixelExpr, invoke PixelOps.pack to pack
them into an int. To extract individual color components from an int, use methods red, green,
and blue in PixelOps. To update individual colors in a pixel, use PixelOps methods setRed,
setGreen, and setBlue.
The image type in PLC Language will be represented using a java.awt.image.BufferedImage.
An image object is instantiated when the declaration is elaborated. An image always has a
size—either it obtains the size from a Dimension in the NameDef in its declaration, or its
declaration has an initializer, and the size is determined from the initializer. If it does not have
either a Dimension or an initializer, throw a CodeGenException. Once instantiated, the size of
the image does not change.
Make sure to add import statements to your generated code for all of the
edu.ufl.cise.cop4020fa23.runtime classes that are used in your generated Java code.
AST Node Code to add to StringBuilder
(shown in red)
_X_ means that visiting _X_ should result in code
being added to represent X.
Program::= Type IDENT NameDef*
Block
public class _IDENT_ {
public static _Type_ apply(
_NameDef*_
) _Block
}
Note: parameters from _NameDef*_ are separated
by commas
Block ::= BlockElem* { _BlockElem*_ }
BlockElem ::= Declaration | Statement Note: Declarations and Statements are terminated
with ;
NameDef ::= Type Dimension?
IDENT _Type_ _name_
Where _name_ is the Java name of the IDENT.
(The dimension will be visited in the parent
Declaration)
Type ::= image | pixel | int | string | void
| boolean
BufferedImage | int | int | String | void | Boolean
Declaration::= NameDef If NameDef.type != IMAGE
_NameDef_
If NameDef.type is image, instantiate a
BufferedImage object using
ImageOps.makeImage.The size of the image
comes from the Dimension object in the NameDef.
If there is no Dimension object, this is an error:
throw a CodeGenException.
(Note: you may choose a different way to divide
the code generation between visitDeclaration and
visitNameDef)
final BufferedImage NameDef.javaName =
ImageOps.makeImage( _Dimension_ )
Declaration::= NameDef Expr If NameDef.type != IMAGE
_NameDef_ = _Expr_
If NameDef.type is an image, there are several
options for the type of the Expr.
If Expr.type is string, then the value should be the
URL of an image which is used to initialize the
declared variable. If the NameDef has a size, then
the image is resized to the given size. Otherwise, it
takes the size of the loaded image.
Use edu.ufl.cise.cop4020fa23.runtime.FileURLIO
readImage (with or without length and width
parameters as appropriate)
If Expr.type is an image and the NameDef does not
have a Dimension, then the image being declared
gets its size from the image on the right side. Use
ImageOps.cloneImage.
If Expr.Type is an image and the NameDef does
have a Dimension, then the image being declared
is initialized to a resized version of the image in the
Expr. Use ImageOps.copyAndResize.
Expr::= ConditionalExpr
| BinaryExpr
| unaryOp Expr
| PostFixExpr
| StringLitExpr _StringLitExpr_.getText
| NumLitExpr _NumLitExpr_.getText
| IdentExpr _IdentExpr_.getNameDef().getJavaName()
ConstExpr If ConsExpr.name = Z then 255
else get hex String literal representing the
RGB representation of the corresponding
java.awt.Color.
Example:
Let the PLC Lang constant be BLUE.
This corresponds to the java Color constant
java.awt.Color.BLUE.
Get the packed pixel version of the color with
getRGB()
Convert to a String with Integer.toHexString
Prepend “0x” to make it a Java hex literal.
Putting it all together, you get
"0x" +
Integer.toHexString(Color.BLUE.getRGB())
Which is
0xff0000ff
| BooleanLitExpr true or false
ExpandedPixelExpr
ConditionalExpr ::= ExprGuardExpr
ExprTrueExpr ExprFalseExpr
( _ ExprGuardExpr_ ? _ ExprTrueExpr _
: _ ExprFalseExpr _ )
BinaryExpr ::= ExprleftExpr op ExprrigthExpr If ExprleftExpr.type is string and op is EQ
_ ExprleftExpr_ .equals( _ ExprrigthExpr _ )
If op is EXP
((int)Math.round(Math.pow( _ ExprleftExpr _ , _
ExprrigthExpr _ )))
Otherwise
(_ ExprleftExpr _ _op_ _ ExprrigthExpr _)
See notes below for handling Pixel and Image
types.
UnaryExpr ::= op Expr ( _op_ _Expr_ )
Note: you do not need to handle width and height
in this assignment
If the op is RES_width or RES_height, we know
that the type is image. Invoke the getWidth or
getHeight methods from BufferedImage on the
image.
If op == RES_height
( __Expr __ .getHeight())
If op == RES_width
( __Expr __ .getWidth())
PostfixExpr::= Expr PixelSelector?
ChannelSelector?
If Expr.type is Pixel
_ChannelSelector_ ( _Expr_ )
Otherwise it is an image
If PixelSelector != null && ChannelSelector ==null
Generate code to get the value of the pixel at the
indicated location.
ImageOps.getRGB( _Expr_ , _PixelSelector _ )
If PixelSelector != null && ChannelSelector != null,
generate code to get the value of the pixel at the
indicated location and to invoke PixelOps.red,
PixelOps.green, or PixelOps.blue. (You may want
to visit the ChannelSelector, passing info that this is
in the context of an expression as indicated here, or
you may want to just get the value from
visitPostfixExpr)
_ChannelSelector_ (ImageOps.getRGB( _Expr_ ,
_PixelSelector_ ))
If PixelSelector == null && ChannelSelector != null,
generate code to invoke the
ImageOPs.extractRed,extractGreen, or extractBlue
method to return a new image containing the
indicated color channel.
ImageOps.extractRed( _Expr_ )
(or extractBlue or extractGreen)
ChannelSelector ::= red | green | blue See PostfixExpr rule for how to handle this in
context of an expression. See LValue for how to
handle in context of an LValue
PixelSelector ::= ExprxExpr ExpryExpr __ExprxExpr__ , __ExpryExpr__
ExpandedPixelExpr ::= Exprred Exprgreen
Exprblue PixelOps.pack(_Exprred_, _Exprgreen_, _Exprblue_)
Dimension ::= Exprwidth Exprheight
__Exprwidth__ , __Exprheight __
LValue ::= IDENT PixelSelector?
ChannelSelector?
_Ident_.getNameDef().getJavaName()
(PixelSelector and ChannelSelector if present, must
be visited. It may be easier to invoke this methods
from the parent AssignmentStatement. )
Statement::=
AssignmentStatement |
WriteStatement |
DoStatement |
IfStatement |
ReturnStatement |
StatementBlock
AssignmentStatement ::= LValue Expr If LValue.varType == image. See explanation
below.
If LValue.varType == pixel and
LValue.ChannelSelector != null
PixelOps.setRed(_LValue_ , _Expr_)
or setGreen or SetBlue
Otherwise
_LValue_ = _Expr_
WriteStatement ::= Expr ConsoleIO.write( _Expr_ )
The ConsoleIO class includes an overloaded
method write for each Java type that represents a
PLC Language type. Thus, you can simply
generate code to call the write method and let the
Java compiler determine which overloaded version
to use. The exception is that int and pixel in PLC
Language are both represented by a Java int.
When the type of Expr is pixel, you need to use the
writePixel method.
DoStatement ::= GuardedBlock+ See below
IfStatement ::= GuardedBlock+ See below
GuardedBlock := Expr Block Handling depends on the context. See description
of DoStatement and IfStatement below.
ReturnStatement ::= Expr return _Expr_
StatementBlock ::= Block _Block_
Binary Expressions with Images and Pixels
Binary operations on these types are generally carried out component-wise.
For example, for images im0 and im1, the expression (e0+e1) yields a new image with the
individual pixels added together componentwise. Pixels in turn are also added componentwise.
For pixels p0, and p1, the expression (p0 + p1) yields a pixel with individual color channel
values added together. Any time the value of a color channel goes out of the range [0,255], it is
truncated to the maximum or minimum value.
When a binary operation has one operand with a “larger” type than the other, the value of the
smaller type is replicated. So for example, im0/2 would divide each pixel in im0 by 2. For each
pixel, each individual color channel value is divided by 2.
Several routines in edu.ufl.cise.cop4020fa23.ImageOps have been provided to implement
binary expressions with images and pixels. See binaryImageIMageOp (combine two images),
binaryImagePixelOp (combine an image with a pixel), binaryImageIntOp (combine an image
with an int), binaryPackedPixelPixelOP (combine two pixels), binaryPackedPixelIntOp (combine
a pixel with an int) and binaryPackedPixelBooleanOP (compare two pixels Boolean operator).
Semantics of IfStatement
You will need to figure out the details yourself.
The semantics are similar to Dijkstra’s guarded command if statement except our version is not
non-deterministic (i.e. is deterministic). The guarded blocks are evaluated starting from the
top. One other difference is that Dijkstra’s version requires that at least one guard be true. In
our version, if none of the guards are true, nothing will happen.
In other words, if the guards are G0, G1, .. Gn, and the corresponding blocks are B0, B1,..,Bn,
Guards are evaluated in turn, starting with G0. When a guard, say Gi, is true, execute the
corresponding Block Bi. That is the end of the if statement. The Java code would look
something like “if (G0) {B0;} else if (G1) {B1;}… else if (Gn) {Bn;}”
Semantics of DoStatement
You will need to figure out the details yourself.
The semantics are like Dijkstra’s guarded command do-od statement except our version is not
non-deterministic (i.e. is deterministic). In each iteration, the guarded blocks are evaluated
starting from the top. (Note that this semantic choice was made for ease of implementation in a
class project. There are alternatives that would probably be more useful in practice.) The loop
terminates when none of the guards are true.
In other words, if the guards are G0, G1, .. Gn, and the corresponding blocks are B0, B1,..,Bn,
Guards are evaluated in turn, starting with G0. When a guard, say Gi is true, execute the
corresponding Block Bi. That is an iteration. Repeat, starting at the top with G0 again, for each
iteration. The statement terminates when none of the guards are true.
Assignment statements where LValue.VarType == IMAGE
There are several cases:
• PixelSelector == null and ChannelSelector == null (i.e. something like im0 = expr; )
• If Expr.type = IMAGE, use ImageOps.copyInto to copy the Expr image into the
LValue image
• If Expr.type = PIXEL, use ImageOps.setAllPixels to update each pixel of the LValue
image with the given pixel value
• If Expr.type = STRING, load the image from the URL in the string, and resizing to the
size of the LValue image. Then copy the pixels of the loaded image into the LValue
image. Use FileURLIO.readImage and ImageOps.copyInto
• ChannelSelector != null (i.e im0:red = expr or im0[x,y]:red = expr)
To simplify the project a little, you may skip this case.
Throw an UnsupportedOperationException instead
• PixelSelector != null and ChannelSelector == null (i.e. something like im[x1,x2] =expr)
This is the most interesting part of the language. Recall that we could write statements like
im0[x,y] = expr where x and/or y are not visible in the current scope. In this case, we added a
SyntheticNameDef object for the variable. For each variable whose nameDef is actually a
SyntheticNameDef object, we generate code for an implicit loop over the values of that variable.
For example, suppose we have a statement
Im0[x,y] = im1[y,x]
where x and y are not previously declared and thus their NameDef objects in the AST are
SyntheticNameDef. Generate code to loop over x and y from 0 to im0.getWidth() and from 0 to
im0.getHeight(), respectively and update each pixel with the value of the expression on the right
side.
See the generated code in the starter Junit test cases for example.
Notes on Junit Test cases
Many of the test cases have an example of generated source code. (Sorry about the
formatting). It is OK if your generated code does not exactly match mine as long as the
semantics are correctly implemented.
The show(BufferedImage) routine will request and wait for console input if WAIT_FOR_INPUT
is true. This prevents the IDE from immediately closing the image frame before you have a
chance to see it. Change the value of WAIT_FOR_INPUT to false if you just want to run all the
test cases as quickly as possible.