ページ更新: 2004-07-11 (日) (5485日前)

JFlex 1.4 付属の FAQ をぼちぼち訳してみるかも知れないので、貼っておく。

[編集]

JFlex - Frequently Asked Questions #

[編集]

古い JLex の仕様書を JFlex で使用できますか? #

Q.Can I use my old JLex specifications with JFlex?

Yes. You usually can use them unchanged. See section [porting from JLex] of the manual for more information on that topic.

はい。通常、それは変更なしに使用できます。マニュアルの「JFlexからの移植」節に、 この件についての詳しい情報があります。

[編集]

JFlex の最新版はどこで入手できますか? #

Q. Where can I get the latest version of JFlex?

Check out the JFlex website at http://www.jflex.de. There is an announcement mailing list, in case you want to be notified of new JFlex releases.

JFlex のWebサイト http://www.jflex.de を調べてください。 また、announcement メーリングリストにて、新しい JFlex がリリースされたときに通知を受けることができます。

[編集]

JFlex はどんなプラットホームをサポートしていますか? #

Q.What platforms does JFlex support?

JFlex is written with Sun's JDK 1.1 and produces JDK 1.1 compatible code. It should run on any platform that supports JRE/JDK 1.1 or above. At the JFlex website, there is a list of platforms JFlex has been positively tested on.

JFlex は Sun の JDK 1.1 で書かれており、JDK 1.1 互換のコードを生成します。 よって、JRE/JDK 1.1 以上の環境であれば、どのプラットホームでも動作するべきです。 JFlex Webサイトにて、JFlex がテストされたプラットホームのリストがあります。

[編集]

大規模なスキャナ(状態が250以上)を使ったプログラムがクラッシュします #

Q.My program crashes when I use a large scanner (about 250 states)

No problem. Just use the %pack directive for more compression of the transition table. See section [code generation] of the manual for more information on this topic.

大丈夫です。 %pack ディレクティブを使えば、遷移表をもっと圧縮できます。 マニュアルの「コード生成」節に詳しい情報があります。

  • 訳注。たぶん、1つのメソッドかクラスが大きくなりすぎて、Javaがクラッシュすることを示していると思われるが、未確認。
[編集]

I get a java.lang.VerifyError when I start my program with java -verify #

This is in fact the same problem as described above. You can use the %pack directive to avoid it. See section [code generation] of the manual for more information on this topic.

[編集]

JFlex と CUP を一緒に使えますか? #

Q. Can I use CUP and JFlex together?

You can. See section [Working together: JFlex and CUP] of the manual on how to do it.

できます。 マニュアルの「一緒に使う:JFlex と CUP」にどうすればよいか載っています。

[編集]

JFlex で生成したコードを商用で使えますか? #

Q. Can I use the generated code of my JFlex specification commercially?

You can use your generated code without restriction. See the file copyright for more information.

[編集]

I want my scanner to read from a network byte stream or from interactive stdin. Can I do this with JFlex? #

This actually depends on the syntax of the input you want to read. The problem is, that for some expressions the scanner needs one character of lookahead to decide which action to execute. For interactive use and network streams this is very inconvenient, because the stream doesn't send an EOF (or any other data) when the user stops typing while the scanner just waits for the next character and doesn't return a symbol. Since version 1.1.1 of JFlex this problem can be avoided because of a little more analysis at generation time. Take a look at these three rules:

1. a
2. a*
3. ";"

When the scanner has read one a, an additional input character is needed to decide, if this matches rule 1 (just one a) or rule two (when the next character is another a). With input aaa the scanner also has to read one additional character, because it is supposed to return the longest match (so if there comes another a, the match is aaaa and not only aaa). But: When the scanner reads a ";", it does not need an additional character and can immediatly execute the action for rule number 3. This is the case for all rules that are not prefixes of any other rules in the specification and that have a fixed length end (so (a* b) is ok but (a b*) is not).

For your application this means: if all commands (or whatever units of input you have) are terminated by some delimiter (for instance ";" or LF or "end") reading from a network bytestream or an interactive stream works fine with JFlex.

[編集]

文字列をスキャナへの入力にするにはどうすればいい? #

Q.How can I let my scanner read its input from a string?

String myString = "some input";
Scanner myScanner = new Scanner( new java.io.StringReader(myString) );
[編集]

Why do standalone scanners have a different standard return type (int instead of Yytoken)? #

That's because int is a predefined type in Java and Yytoken is not. If a scanner generated with %standalone option would have return type Yytoken, you would have to provide this class for every standalone scanner you write. In most cases you don't want to do that, because the scanner wouldn't be really standalone then. The standard Yytoken for non standalone cases stems from JLex and is only kept for compatibility (it's rarely used anyway). If you still really want Yytoken as return type in a standalone scanner, you can always explicitly specify it with %type Yytoken. If you just want to test your scanner scanner and see what it does without a parser attached, use %debug instead of %standalone.

[編集]

The expression .*$ used to work, but now JFlex complains about trailing context #

This is because .*$ is now matched with the more general lookahead operator / and is equivalent to .* / \r|\n|\r\n. The problem with this expression is that the character class . contains the \r character and therefore matches the beginning of the trailing context. Earlier versions of JFlex used a special implementation for $ and had no lookahead operator.

A simple workaround is not to use the . class, but to use the expression [^\r\n] instead (so the full expression is now [^\r\n]*$ and should work without error).

[編集]

I use %8bit and get an Exception, but I know my platform only uses 8 bit. Is %8bit broken? #

Short answer: not broken, use %unicode. Long answer: Most probably this is an encoding problem. Java uses Unicode internally and converts the bytes it reads from files (or somewhere else) to Unicode first. The 8 bit value of your platform may not be 8 bit anymore when converted to Unicode. On many Windozes for instance Cp1252 (Windows-Latin-1) is used as standard encoding, and there the character "single right quotation mark" has code \x92 but after conversion to Unicode it's \u2019 which is not 8 bit any more. See also the section on Encodings, platforms, and Unicode of the JFlex manual for more information.

[編集]

My scanner needs to read a file that is not in my platforms standard encoding, but in encoding XYZ. How? #

Since the scanner reads Java Unicode characters, it is independent of the actual character encoding a file or a string uses. The transformation byte-stream to Java characters for files usually happens in the java.io.InputStreamReader object connected with the input stream. Class java.io.FileReader uses the platforms default encoding automatically. If you would like to explicitly specify another encoding, for instance UTF-8, you could do something like

Reader r = new InputStreamReader(new FileInputStream(file), "UTF8");

Now you have a Reader r that can be passed to the scanner's constructor in the usual way. For more information on encodings see also Sun's JDK documentation, especially in Guide to Features - Java Platform item Internationalization and there the FAQ and Supported Encodings.

[編集]

JFlex generates with 0 errors, 0 warnings, but the generated code does not compile #

If you are sure that the problem is because of JFlex and not because of user code in the specification, try to run JFlex without the JIT compiler. Some versions of Sun's JIT compiler sunwjit in JDK 1.2.2 on Solaris 7/Intel seem to have a bug that is triggered by JFlex (if your scanner compiles ok, there should be no problem).

Please do report this as a bug to lsf@jflex.de in any case.

If there is a problem with a specific platform/JDK/JIT combination it may be important for other people to know. Also, I might be able to work around the JIT bug from the JFlex side.