Bug 74126 - Compiler should support new hexadecimal floating-point literals
Summary: Compiler should support new hexadecimal floating-point literals
Status: VERIFIED FIXED
Alias: None
Product: JDT
Classification: Eclipse Project
Component: Core (show other bugs)
Version: 3.0   Edit
Hardware: PC Windows XP
: P3 normal (vote)
Target Milestone: 3.1 M3   Edit
Assignee: Olivier Thomann CLA
QA Contact:
URL:
Whiteboard:
Keywords:
Depends on:
Blocks:
 
Reported: 2004-09-16 17:29 EDT by Olivier Thomann CLA
Modified: 2004-11-02 11:57 EST (History)
2 users (show)

See Also:

Attachments
Apply on HEAD (1.85 KB, application/octet-stream)
2004-09-16 17:37 EDT, Olivier Thomann CLA 		no flags Details
Apply on FloatLiteral and DoubleLiteral (2.75 KB, application/octet-stream)
2004-09-16 18:44 EDT, Olivier Thomann CLA 		no flags Details
Patch to o.e.jdt.core project (26.56 KB, patch)
2004-09-27 17:04 EDT, Jim des Rivieres CLA 		no flags Details | Diff
Patch to o.e.jdt.core.tests.compiler (15.12 KB, patch)
2004-09-27 17:08 EDT, Jim des Rivieres CLA 		no flags Details | Diff
View All Add an attachment (proposed patch, testcase, etc.)

Note You need to log in before you can comment on or make changes to this bug.
Description Olivier Thomann CLA 2004-09-16 17:29:59 EDT 
The JLS3 states that the floating-point literals can now have a binary exponent.

Here are the rules to create a floating-point literal:

FloatingPointLiteral
	DecimalFloatingPointLiteral
	HexadecimalFloatingPointLiteral

DecimalFloatingPointLiteral:
	Digits . Digitsopt ExponentPartopt FloatTypeSuffixopt
	. Digits ExponentPartopt FloatTypeSuffixopt
	Digits ExponentPart FloatTypeSuffixopt
	Digits ExponentPartopt FloatTypeSuffix

ExponentPart:
	ExponentIndicator SignedInteger

ExponentIndicator: one of
	e E

SignedInteger:
	Signopt Digits

Sign: one of
	+ -

FloatTypeSuffix: one of
	f F d D

HexadecimalFloatingPointLiteral:
	HexSignificand BinaryExponent FloatTypeSuffixopt

HexSignificand:
	HexNumeral
	HexNumeral .
	0x HexDigitsopt . HexDigits
	0X HexDigitsopt . HexDigits

BinaryExponent:
	BinaryExponentIndicator SignedInteger

BinaryExponentIndicator: one of
	p P
	
HexNumeral:
	0 x HexDigits
	0 X HexDigits
	
HexDigits:
	HexDigit
	HexDigit HexDigits
	
HexDigit: one of
	0 1 2 3 4 5 6 7 8 9 a b c d e f A B C D E F

The unsupported format is HexadecimalFloatingPointLiteral.
Comment 1 Olivier Thomann CLA 2004-09-16 17:37:42 EDT 
Created attachment 14584 [details]
Apply on HEAD

Here is a zipped file that contains a patch for the scanner, problem reporter,
IProblem and messages.properties.
It seems to work fine. I will write more tests.

The problem remains that the compiler requires 1.5 libraries in order to
compute the new floating-point literal constants.
Comment 2 Olivier Thomann CLA 2004-09-16 18:44:58 EDT 
Created attachment 14586 [details]
Apply on FloatLiteral and DoubleLiteral

Apply the two patches on FloatLiteral and DoubleLiteral. It is a hack to
compute the float or double constant when the compiler runs on 1.4 libraries.
Comment 3 Olivier Thomann CLA 2004-09-16 18:47:00 EDT 
I forgot to change the FloatLiteral class.
Replace:

	try {
		computedValue = Float.valueOf(String.valueOf(source));
	} catch (NumberFormatException e) {
		computedValue = getFloatingPoint();
	} 

with:
	try {
		computedValue = Float.valueOf(String.valueOf(source));
	} catch (NumberFormatException e) {
		try {
			computedValue = getFloatingPoint();
		} catch(NumberFormatException e1) {
			return;
		}
	}
Comment 4 Olivier Thomann CLA 2004-09-20 12:34:21 EDT 
First draft released.
We will review the constant computation in case the libraries used to compile
are < 1.5.
Comment 5 Olivier Thomann CLA 2004-09-20 12:34:55 EDT 
I will add scanner tests to be sure that the new float and double literals are
propertly scanned.
Comment 6 Olivier Thomann CLA 2004-09-20 13:05:36 EDT 
This change requires changes in the ToolFactory.
We have indeed three modes:
- compliance 1.5 and source < 1.5:
=> This is one token
- compliance 1.4 and source < 1.5:
=> This is mode than one token
- compliance 1.5 and source 1.5:
=> This is one token
Comment 7 Olivier Thomann CLA 2004-09-21 13:50:08 EDT 
Fixed and released in HEAD.
Regression tests added.
Comment 8 David Audel CLA 2004-09-24 06:35:32 EDT 
Verified in I200409231635.
Comment 9 Jim des Rivieres CLA 2004-09-27 16:58:24 EDT 
Reopening to include patches for improved handling of hex floating point 
literals.
Comment 10 Jim des Rivieres CLA 2004-09-27 17:04:50 EDT 
Created attachment 14808 [details]
Patch to o.e.jdt.core project

The patch replaces Util.getFloatingPoint(char[]) with
FloatUtil.valueOfHexDoubleLiteral(char[]) and
FloatUtil.valueOfHexFloatLiteral(char[]). These new methods handle all the
corner cases, including rounding, overflow, and gradual underflow.
Comment 11 Jim des Rivieres CLA 2004-09-27 17:08:49 EDT 
Created attachment 14809 [details]
Patch to o.e.jdt.core.tests.compiler

Patch adds a new test, InternalHexFloatTest, to test the new internal FloatUtil

class's methods.
Comment 12 Jim des Rivieres CLA 2004-09-27 17:09:20 EDT 
Reopening to include patches for improved handling of hex floating point 
literals.
Comment 13 Olivier Thomann CLA 2004-09-28 17:45:56 EDT 
Fixed and released in HEAD.
Regression tests added.
Comment 14 David Audel CLA 2004-11-02 11:57:31 EST 
Verified for 3.1M3 with build I20041101