[PATCH] userdiff: add Julia to supported userdiff languages

[PATCH] userdiff: add Julia to supported userdiff languages

[Ryan Zoeller via GitGitGadget]

Add xfuncname and word_regex patterns for Julia[1],
which is a language used in numerical analysis and
computational science.

The default behavior for xfuncname did not allow
functions to be indented, nor functions to have a
macro applied, such as @inline or @generated.

[1]: https://julialang.org

Signed-off-by: Ryan Zoeller <rtzoeller@rtzoeller.com>
---
    userdiff: add Julia to supported userdiff languages
    
    Add xfuncname and word_regex patterns for Julia1 [https://julialang.org]
    , which is a language used in numerical analysis and computational
    science.
    
    The default behavior for xfuncname did not allow functions to be
    indented, nor functions to have a macro applied, such as @inline or
    @generated.
    
    Signed-off-by: Ryan Zoeller rtzoeller@rtzoeller.com
    [rtzoeller@rtzoeller.com]

Published-As: https://github.com/gitgitgadget/git/releases/tag/pr-521%2Frtzoeller%2Fjulia_userdiff-v1
Fetch-It-Via: git fetch https://github.com/gitgitgadget/git pr-521/rtzoeller/julia_userdiff-v1
Pull-Request: https://github.com/gitgitgadget/git/pull/521

 Documentation/gitattributes.txt |  2 ++
 t/t4018-diff-funcname.sh        |  1 +
 t/t4018/julia-function          |  5 +++++
 t/t4018/julia-indented-function |  8 ++++++++
 t/t4018/julia-inline-function   |  5 +++++
 t/t4018/julia-macro             |  5 +++++
 t/t4018/julia-mutable-struct    |  5 +++++
 t/t4018/julia-struct            |  5 +++++
 userdiff.c                      | 15 +++++++++++++++
 9 files changed, 51 insertions(+)
 create mode 100644 t/t4018/julia-function
 create mode 100644 t/t4018/julia-indented-function
 create mode 100644 t/t4018/julia-inline-function
 create mode 100644 t/t4018/julia-macro
 create mode 100644 t/t4018/julia-mutable-struct
 create mode 100644 t/t4018/julia-struct

diff --git a/Documentation/gitattributes.txt b/Documentation/gitattributes.txt
index 508fe713c4..d39dc727e3 100644
--- a/Documentation/gitattributes.txt
+++ b/Documentation/gitattributes.txt
@@ -824,6 +824,8 @@ patterns are available:
 
 - `java` suitable for source code in the Java language.
 
+- `julia` suitable for source code in the Julia language.
+
 - `matlab` suitable for source code in the MATLAB and Octave languages.
 
 - `objc` suitable for source code in the Objective-C language.
diff --git a/t/t4018-diff-funcname.sh b/t/t4018-diff-funcname.sh
index c0f4839543..d4613eb7d2 100755
--- a/t/t4018-diff-funcname.sh
+++ b/t/t4018-diff-funcname.sh
@@ -38,6 +38,7 @@ diffpatterns="
 	golang
 	html
 	java
+	julia
 	matlab
 	objc
 	pascal
diff --git a/t/t4018/julia-function b/t/t4018/julia-function
new file mode 100644
index 0000000000..a2eab83c27
--- /dev/null
+++ b/t/t4018/julia-function
@@ -0,0 +1,5 @@
+function RIGHT()
+    # A comment
+    # Another comment
+    return ChangeMe
+end
diff --git a/t/t4018/julia-indented-function b/t/t4018/julia-indented-function
new file mode 100644
index 0000000000..2d48aabcdb
--- /dev/null
+++ b/t/t4018/julia-indented-function
@@ -0,0 +1,8 @@
+function outer_function()
+    function RIGHT()
+        for i = 1:10
+            print(i)
+        end
+        # ChangeMe
+    end
+end
diff --git a/t/t4018/julia-inline-function b/t/t4018/julia-inline-function
new file mode 100644
index 0000000000..5806f224fb
--- /dev/null
+++ b/t/t4018/julia-inline-function
@@ -0,0 +1,5 @@
+@inline function RIGHT()
+    # Prints Hello, then something else.
+    println("Hello")
+    println("ChangeMe")
+end
diff --git a/t/t4018/julia-macro b/t/t4018/julia-macro
new file mode 100644
index 0000000000..1d18bc2750
--- /dev/null
+++ b/t/t4018/julia-macro
@@ -0,0 +1,5 @@
+macro RIGHT()
+    # First comment
+    # Second comment
+    return :( println("ChangeMe") )
+end
diff --git a/t/t4018/julia-mutable-struct b/t/t4018/julia-mutable-struct
new file mode 100644
index 0000000000..db82017ba0
--- /dev/null
+++ b/t/t4018/julia-mutable-struct
@@ -0,0 +1,5 @@
+mutable struct RIGHT
+    x
+    y::Int
+    ChangeMe
+end
diff --git a/t/t4018/julia-struct b/t/t4018/julia-struct
new file mode 100644
index 0000000000..d3d2bda8cb
--- /dev/null
+++ b/t/t4018/julia-struct
@@ -0,0 +1,5 @@
+struct RIGHT
+    x
+    y::Int
+    ChangeMe
+end
diff --git a/userdiff.c b/userdiff.c
index efbe05e5a5..b5e938b1c2 100644
--- a/userdiff.c
+++ b/userdiff.c
@@ -79,6 +79,21 @@ PATTERNS("java",
 	 "|[-+0-9.e]+[fFlL]?|0[xXbB]?[0-9a-fA-F]+[lL]?"
 	 "|[-+*/<>%&^|=!]="
 	 "|--|\\+\\+|<<=?|>>>?=?|&&|\\|\\|"),
+PATTERNS("julia",
+	 "^[ \t]*(((mutable[ \t]+)?struct|(@.+[ \t])?function|macro)[ \t].*)$",
+	 /* -- */
+	 /* Binary literals */
+	 "[-+]?0b[01]+"
+	 /* Hexadecimal literals */
+	 "|[-+]?0x[0-9a-fA-F]+"
+	 /* Real and complex literals */
+	 "|[-+0-9.e_(im)]+"
+	 /* Should theoretically allow Unicode characters as part of
+	  * a word, such as U+2211. However, Julia reserves most of the
+	  * U+2200-U+22FF range (as well as others) as user-defined operators,
+	  * therefore they are not handled in this regex. */
+	 "|[a-zA-Z_][a-zA-Z0-9_!]*"
+	 "|--|\\+\\+|<<=?|>>>=?|>>=?|\\\\\\\\=?|//=?|&&|\\|\\||::|->|[-+*/<>%^&|=!$]=?"),
 PATTERNS("matlab",
 	 /*
 	  * Octave pattern is mostly the same as matlab, except that '%%%' and

base-commit: 042ed3e048af08014487d19196984347e3be7d1c
-- 
gitgitgadget

Re: [PATCH] userdiff: add Julia to supported userdiff languages

[Johannes Schindelin]

Hi,

On Fri, 10 Jan 2020, Ryan Zoeller via GitGitGadget wrote:

> Add xfuncname and word_regex patterns for Julia[1],
> which is a language used in numerical analysis and
> computational science.
>
> The default behavior for xfuncname did not allow
> functions to be indented, nor functions to have a
> macro applied, such as @inline or @generated.
>
> [1]: https://julialang.org
>
> Signed-off-by: Ryan Zoeller <rtzoeller@rtzoeller.com>
> ---
>     userdiff: add Julia to supported userdiff languages
>
>     Add xfuncname and word_regex patterns for Julia1 [https://julialang.org]
>     , which is a language used in numerical analysis and computational
>     science.
>
>     The default behavior for xfuncname did not allow functions to be
>     indented, nor functions to have a macro applied, such as @inline or
>     @generated.
>
>     Signed-off-by: Ryan Zoeller rtzoeller@rtzoeller.com
>     [rtzoeller@rtzoeller.com]

Sorry about that. In my recent work to fold in the cover letter into
single-patch contributions, it was mentioned that this could come back to
bite us: By default, GitHub uses the commit message of single-commit PRs
as PR description, and if contributors do not change that, it essentially
repeats the commit message.

Sadly, I won't be able to justify working even more on GitGitGadget this
week (it took a sizable chunk out of my time budget and I have to make up
for that first).

Ciao,
Johannes

>
> Published-As: https://github.com/gitgitgadget/git/releases/tag/pr-521%2Frtzoeller%2Fjulia_userdiff-v1
> Fetch-It-Via: git fetch https://github.com/gitgitgadget/git pr-521/rtzoeller/julia_userdiff-v1
> Pull-Request: https://github.com/gitgitgadget/git/pull/521
>
>  Documentation/gitattributes.txt |  2 ++
>  t/t4018-diff-funcname.sh        |  1 +
>  t/t4018/julia-function          |  5 +++++
>  t/t4018/julia-indented-function |  8 ++++++++
>  t/t4018/julia-inline-function   |  5 +++++
>  t/t4018/julia-macro             |  5 +++++
>  t/t4018/julia-mutable-struct    |  5 +++++
>  t/t4018/julia-struct            |  5 +++++
>  userdiff.c                      | 15 +++++++++++++++
>  9 files changed, 51 insertions(+)
>  create mode 100644 t/t4018/julia-function
>  create mode 100644 t/t4018/julia-indented-function
>  create mode 100644 t/t4018/julia-inline-function
>  create mode 100644 t/t4018/julia-macro
>  create mode 100644 t/t4018/julia-mutable-struct
>  create mode 100644 t/t4018/julia-struct
>
> diff --git a/Documentation/gitattributes.txt b/Documentation/gitattributes.txt
> index 508fe713c4..d39dc727e3 100644
> --- a/Documentation/gitattributes.txt
> +++ b/Documentation/gitattributes.txt
> @@ -824,6 +824,8 @@ patterns are available:
>
>  - `java` suitable for source code in the Java language.
>
> +- `julia` suitable for source code in the Julia language.
> +
>  - `matlab` suitable for source code in the MATLAB and Octave languages.
>
>  - `objc` suitable for source code in the Objective-C language.
> diff --git a/t/t4018-diff-funcname.sh b/t/t4018-diff-funcname.sh
> index c0f4839543..d4613eb7d2 100755
> --- a/t/t4018-diff-funcname.sh
> +++ b/t/t4018-diff-funcname.sh
> @@ -38,6 +38,7 @@ diffpatterns="
>  	golang
>  	html
>  	java
> +	julia
>  	matlab
>  	objc
>  	pascal
> diff --git a/t/t4018/julia-function b/t/t4018/julia-function
> new file mode 100644
> index 0000000000..a2eab83c27
> --- /dev/null
> +++ b/t/t4018/julia-function
> @@ -0,0 +1,5 @@
> +function RIGHT()
> +    # A comment
> +    # Another comment
> +    return ChangeMe
> +end
> diff --git a/t/t4018/julia-indented-function b/t/t4018/julia-indented-function
> new file mode 100644
> index 0000000000..2d48aabcdb
> --- /dev/null
> +++ b/t/t4018/julia-indented-function
> @@ -0,0 +1,8 @@
> +function outer_function()
> +    function RIGHT()
> +        for i = 1:10
> +            print(i)
> +        end
> +        # ChangeMe
> +    end
> +end
> diff --git a/t/t4018/julia-inline-function b/t/t4018/julia-inline-function
> new file mode 100644
> index 0000000000..5806f224fb
> --- /dev/null
> +++ b/t/t4018/julia-inline-function
> @@ -0,0 +1,5 @@
> +@inline function RIGHT()
> +    # Prints Hello, then something else.
> +    println("Hello")
> +    println("ChangeMe")
> +end
> diff --git a/t/t4018/julia-macro b/t/t4018/julia-macro
> new file mode 100644
> index 0000000000..1d18bc2750
> --- /dev/null
> +++ b/t/t4018/julia-macro
> @@ -0,0 +1,5 @@
> +macro RIGHT()
> +    # First comment
> +    # Second comment
> +    return :( println("ChangeMe") )
> +end
> diff --git a/t/t4018/julia-mutable-struct b/t/t4018/julia-mutable-struct
> new file mode 100644
> index 0000000000..db82017ba0
> --- /dev/null
> +++ b/t/t4018/julia-mutable-struct
> @@ -0,0 +1,5 @@
> +mutable struct RIGHT
> +    x
> +    y::Int
> +    ChangeMe
> +end
> diff --git a/t/t4018/julia-struct b/t/t4018/julia-struct
> new file mode 100644
> index 0000000000..d3d2bda8cb
> --- /dev/null
> +++ b/t/t4018/julia-struct
> @@ -0,0 +1,5 @@
> +struct RIGHT
> +    x
> +    y::Int
> +    ChangeMe
> +end
> diff --git a/userdiff.c b/userdiff.c
> index efbe05e5a5..b5e938b1c2 100644
> --- a/userdiff.c
> +++ b/userdiff.c
> @@ -79,6 +79,21 @@ PATTERNS("java",
>  	 "|[-+0-9.e]+[fFlL]?|0[xXbB]?[0-9a-fA-F]+[lL]?"
>  	 "|[-+*/<>%&^|=!]="
>  	 "|--|\\+\\+|<<=?|>>>?=?|&&|\\|\\|"),
> +PATTERNS("julia",
> +	 "^[ \t]*(((mutable[ \t]+)?struct|(@.+[ \t])?function|macro)[ \t].*)$",
> +	 /* -- */
> +	 /* Binary literals */
> +	 "[-+]?0b[01]+"
> +	 /* Hexadecimal literals */
> +	 "|[-+]?0x[0-9a-fA-F]+"
> +	 /* Real and complex literals */
> +	 "|[-+0-9.e_(im)]+"
> +	 /* Should theoretically allow Unicode characters as part of
> +	  * a word, such as U+2211. However, Julia reserves most of the
> +	  * U+2200-U+22FF range (as well as others) as user-defined operators,
> +	  * therefore they are not handled in this regex. */
> +	 "|[a-zA-Z_][a-zA-Z0-9_!]*"
> +	 "|--|\\+\\+|<<=?|>>>=?|>>=?|\\\\\\\\=?|//=?|&&|\\|\\||::|->|[-+*/<>%^&|=!$]=?"),
>  PATTERNS("matlab",
>  	 /*
>  	  * Octave pattern is mostly the same as matlab, except that '%%%' and
>
> base-commit: 042ed3e048af08014487d19196984347e3be7d1c
> --
> gitgitgadget
>

Re: [PATCH] userdiff: add Julia to supported userdiff languages

[Johannes Sixt]

Am 10.01.20 um 04:10 schrieb Ryan Zoeller via GitGitGadget:
> Add xfuncname and word_regex patterns for Julia[1],
> which is a language used in numerical analysis and
> computational science.
> 
> The default behavior for xfuncname did not allow
> functions to be indented, nor functions to have a
> macro applied, such as @inline or @generated.
> 
> [1]: https://julialang.org
> 
> Signed-off-by: Ryan Zoeller <rtzoeller@rtzoeller.com>
> ---
> 
> Published-As: https://github.com/gitgitgadget/git/releases/tag/pr-521%2Frtzoeller%2Fjulia_userdiff-v1
> Fetch-It-Via: git fetch https://github.com/gitgitgadget/git pr-521/rtzoeller/julia_userdiff-v1
> Pull-Request: https://github.com/gitgitgadget/git/pull/521
> 
>  Documentation/gitattributes.txt |  2 ++
>  t/t4018-diff-funcname.sh        |  1 +
>  t/t4018/julia-function          |  5 +++++
>  t/t4018/julia-indented-function |  8 ++++++++
>  t/t4018/julia-inline-function   |  5 +++++
>  t/t4018/julia-macro             |  5 +++++
>  t/t4018/julia-mutable-struct    |  5 +++++
>  t/t4018/julia-struct            |  5 +++++
>  userdiff.c                      | 15 +++++++++++++++
>  9 files changed, 51 insertions(+)
>  create mode 100644 t/t4018/julia-function
>  create mode 100644 t/t4018/julia-indented-function
>  create mode 100644 t/t4018/julia-inline-function
>  create mode 100644 t/t4018/julia-macro
>  create mode 100644 t/t4018/julia-mutable-struct
>  create mode 100644 t/t4018/julia-struct

The tests all look good.

> diff --git a/userdiff.c b/userdiff.c
> index efbe05e5a5..b5e938b1c2 100644
> --- a/userdiff.c
> +++ b/userdiff.c
> @@ -79,6 +79,21 @@ PATTERNS("java",
>  	 "|[-+0-9.e]+[fFlL]?|0[xXbB]?[0-9a-fA-F]+[lL]?"
>  	 "|[-+*/<>%&^|=!]="
>  	 "|--|\\+\\+|<<=?|>>>?=?|&&|\\|\\|"),
> +PATTERNS("julia",
> +	 "^[ \t]*(((mutable[ \t]+)?struct|(@.+[ \t])?function|macro)[ \t].*)$",

Looks good to me.

> +	 /* -- */
> +	 /* Binary literals */
> +	 "[-+]?0b[01]+"
> +	 /* Hexadecimal literals */
> +	 "|[-+]?0x[0-9a-fA-F]+"

These two could be merged into

	/* Binary and hexadecimal literals */
	"|0[bx][0-9a-fA-F]+"

Note that I did not insert [-+]? at the front. Even though most if not
all patterns allow a sign, they are usually wrong to do so, because they
misclassify a change from 'a+1' to 'a+2' as 'a[-+1-]{++2+}' instead of
the correct 'a+[-1-]{+2+}'.

> +	 /* Real and complex literals */
> +	 "|[-+0-9.e_(im)]+"

I am curious: is '(1+2i)' a single literal -- including the parentheses?
The expression would also mistake the character sequence '-1)+(2+' as a
single word; is it intended?

> +	 /* Should theoretically allow Unicode characters as part of
> +	  * a word, such as U+2211. However, Julia reserves most of the
> +	  * U+2200-U+22FF range (as well as others) as user-defined operators,
> +	  * therefore they are not handled in this regex. */
> +	 "|[a-zA-Z_][a-zA-Z0-9_!]*"
> +	 "|--|\\+\\+|<<=?|>>>=?|>>=?|\\\\\\\\=?|//=?|&&|\\|\\||::|->|[-+*/<>%^&|=!$]=?"),

The last sub-expression permits single-character operators in addition
to their forms with a '=' appended (computing assignment, I presume).
You could remove the trailing ? because single non-whitespace characters
are always a word of their own, even if they are not caught by the word
regexp.

>  PATTERNS("matlab",
>  	 /*
>  	  * Octave pattern is mostly the same as matlab, except that '%%%' and
> 
> base-commit: 042ed3e048af08014487d19196984347e3be7d1c
> 

-- Hannes

Re: [PATCH] userdiff: add Julia to supported userdiff languages

[Ryan Zoeller]

On Friday, January 10, 2020 11:43 AM, Johannes Sixt <j6t@kdbg.org> wrote:

> Am 10.01.20 um 04:10 schrieb Ryan Zoeller via GitGitGadget:
>
> > Add xfuncname and word_regex patterns for Julia1,
> > which is a language used in numerical analysis and
> > computational science.
> > The default behavior for xfuncname did not allow
> > functions to be indented, nor functions to have a
> > macro applied, such as @inline or @generated.
> >
> > Signed-off-by: Ryan Zoeller rtzoeller@rtzoeller.com
> >
> > ----------------------------------------------------
> >
> > Published-As: https://github.com/gitgitgadget/git/releases/tag/pr-521%2Frtzoeller%2Fjulia_userdiff-v1
> > Fetch-It-Via: git fetch https://github.com/gitgitgadget/git pr-521/rtzoeller/julia_userdiff-v1
> > Pull-Request: https://github.com/gitgitgadget/git/pull/521
> > Documentation/gitattributes.txt | 2 ++
> > t/t4018-diff-funcname.sh | 1 +
> > t/t4018/julia-function | 5 +++++
> > t/t4018/julia-indented-function | 8 ++++++++
> > t/t4018/julia-inline-function | 5 +++++
> > t/t4018/julia-macro | 5 +++++
> > t/t4018/julia-mutable-struct | 5 +++++
> > t/t4018/julia-struct | 5 +++++
> > userdiff.c | 15 +++++++++++++++
> > 9 files changed, 51 insertions(+)
> > create mode 100644 t/t4018/julia-function
> > create mode 100644 t/t4018/julia-indented-function
> > create mode 100644 t/t4018/julia-inline-function
> > create mode 100644 t/t4018/julia-macro
> > create mode 100644 t/t4018/julia-mutable-struct
> > create mode 100644 t/t4018/julia-struct
>
> The tests all look good.
>
> > diff --git a/userdiff.c b/userdiff.c
> > index efbe05e5a5..b5e938b1c2 100644
> > --- a/userdiff.c
> > +++ b/userdiff.c
> > @@ -79,6 +79,21 @@ PATTERNS("java",
> > "|[-+0-9.e]+[fFlL]?|0[xXbB]?[0-9a-fA-F]+[lL]?"
> > "|[-+*/<>%&^|=!]="
> > "|--|\\+\\+|<<=?|>>>?=?|&&|\\|\\|"),
> > +PATTERNS("julia",
> >
> > -   "^[ \t](((mutable[ \t]+)?struct|(@.+[ \t])?function|macro)[ \t].)$",
>
> Looks good to me.
>
> > -   /* -- */
> > -   /* Binary literals */
> > -   "[-+]?0b[01]+"
> > -   /* Hexadecimal literals */
> > -   "|[-+]?0x[0-9a-fA-F]+"
>
> These two could be merged into
>
> /* Binary and hexadecimal literals */
> "|0[bx][0-9a-fA-F]+"

I was trying to avoid `0b` being followed by hex characters from being recognized, e.g. 0bFF. This is admittedly not really a concern, so I'm fine making this change to simplify the regular expression.

>
> Note that I did not insert [-+]? at the front. Even though most if not
> all patterns allow a sign, they are usually wrong to do so, because they
> misclassify a change from 'a+1' to 'a+2' as 'a[-+1-]{++2+}' instead of
> the correct 'a+[-1-]{+2+}'.

I'm fine dropping the leading `[-+]?`.

>
> > -   /* Real and complex literals */
> > -   "|[-+0-9.e_(im)]+"
>
> I am curious: is '(1+2i)' a single literal -- including the parentheses?
> The expression would also mistake the character sequence '-1)+(2+' as a
> single word; is it intended?

This part of the regular expression has a pretty major mistake due to me misunderstanding how the parentheses were being interpreted. It should be something along the lines of `([-+0-9.e_]|im)+`.

Julia uses `im` as the designation for an imaginary value; this regex was intended to admit e.g. 1+2im, in addition other numeric values such as 1_000_000 and 1e10.

>
> > -   /* Should theoretically allow Unicode characters as part of
> > -   -   a word, such as U+2211. However, Julia reserves most of the
> > -   -   U+2200-U+22FF range (as well as others) as user-defined operators,
> > -   -   therefore they are not handled in this regex. */
> > -   "|[a-zA-Z_][a-zA-Z0-9_!]*"
> > -   "|--|\\+\\+|<<=?|>>>=?|>>=?|\\\\\\\\=?|//=?|&&|\\|\\||::|->|[-+*/<>%^&|=!$]=?"),
>
> The last sub-expression permits single-character operators in addition
> to their forms with a '=' appended (computing assignment, I presume).
> You could remove the trailing ? because single non-whitespace characters
> are always a word of their own, even if they are not caught by the word
> regexp.

Agreed, I'll drop the trailing ?.

>
> > PATTERNS("matlab",
> > /*
> > * Octave pattern is mostly the same as matlab, except that '%%%' and
> > base-commit: 042ed3e048af08014487d19196984347e3be7d1c
>
> -- Hannes

Thanks for the feedback,
Ryan Zoeller

Re: [PATCH] userdiff: add Julia to supported userdiff languages

[Johannes Sixt]

Am 10.01.20 um 19:15 schrieb Ryan Zoeller:
> On Friday, January 10, 2020 11:43 AM, Johannes Sixt <j6t@kdbg.org> wrote:
>> Am 10.01.20 um 04:10 schrieb Ryan Zoeller via GitGitGadget:
>>> -   /* Real and complex literals */
>>> -   "|[-+0-9.e_(im)]+"
>>
>> I am curious: is '(1+2i)' a single literal -- including the parentheses?
>> The expression would also mistake the character sequence '-1)+(2+' as a
>> single word; is it intended?
> 
> This part of the regular expression has a pretty major mistake due
> to me misunderstanding how the parentheses were being interpreted.
> It should be something along the lines of `([-+0-9.e_]|im)+`.
> 
> Julia uses `im` as the designation for an imaginary value; this regex
> was intended to admit e.g. 1+2im, in addition other numeric values
> such as 1_000_000 and 1e10.
I see. I suggest to treat 1+2im as three words '1', '+', and '2im', and
to model numbers in this way:

	|[0-9][0-9_.]*(e[-+]?[0-9_]*)?(im)?

In particular, require a digit at the begin, and do not allow '-' and
'+' an arbitrary number of times, because it would catch 1+2+3+4 as a
single word.

-- Hannes