When does whitespace in JavaScript matter?

[Peter Michaux]

Hi,

I'm thinking about code minimization. I can think of a few places where
whitespace matters

a + ++b
a++ + b
a - --b
a-- -b
when a line ends without a semi-colon in which case the new line
character matters.

Any others?

Thanks,
Peter

[Randy Webb]

Peter Michaux said the following on 1/8/2007 2:58 PM:

Hi,

I'm thinking about code minimization. I can think of a few places where
whitespace matters

In those 4 instances where whitespace will matter I always wrap it in
parentheses to remove the whitespace issue.

a + ++b
a++ + b
a - --b
a-- -b
when a line ends without a semi-colon in which case the new line
character matters.

That is not always true though:

function someFunction()
{

The newline character there is irrelevant. But, if you replace all new
lines with ;\n then you break the code. Line feeds and semicolons are a
spot where you have to either do it manually or write a JS parser.

Another example where the newline can't be summarily replaced but the
newline doesn't matter is in a loop:

for (i in something)
while(yourWifeI sntPregnant)

Whether the { is on the same line or the next line doesn't matter, the
newline is still irrelevant yet there is no semicolon at the end of it.

You can't summarily say "don't put a semicolon after )" either. Think of
anonymous functions....()
You could check ) to see if it is () or ) but it can still get you in
trouble with a function with no parameter:

function myFunction()
{

}

Where the newline character can't be replaced with a ;newline but if
it's a global anonymous function and you want to remove newlines then
you *must* have the semicolon or it breaks the code.

var x = (function(){
//some code here
})();

Can't remove the semicolon and the newline, one has to be there.

Semicolons and newlines will become your new nightmare :)

--
Randy
Chance Favors The Prepared Mind
comp.lang.javas cript FAQ - http://jibbering.com/faq/index.html
Javascript Best Practices - http://www.JavascriptToolbox.com/bestpractices/

[Duncan Booth]

"Peter Michaux" <pe**********@g mail.comwrote:

Hi,

I'm thinking about code minimization. I can think of a few places where
whitespace matters

a + ++b
a++ + b
a - --b
a-- -b
when a line ends without a semi-colon in which case the new line
character matters.

The new line character only matters some of the time. Unfortunately you
aren't going to know whether or not it matters without doing a full parser.

>
Any others?

A keyword followed by a letter or digit: "return 42;" versus "return42;"

Inside strings and regular expressions.

A comment which immediately follows a division:
a = x / /*this had better not be zero: */y;
z = 1;
versus:
a = x //*this had better not be zero: */y;
z = 1;

[Peter Michaux]

Randy Webb wrote:

Peter Michaux said the following on 1/8/2007 2:58 PM:

Hi,

I'm thinking about code minimization. I can think of a few places where
whitespace matters

<snip>

Semicolons and newlines will become your new nightmare :)

I imagine so. I may just decide to leave new lines in place.

It's more the comments, blank lines and beginning or mid-line spaces
that are a bigger concern. I have been using jslint to check for
problems before jsmin but jslint has many programmer preferences built
in and not just outright problems that will be harmful when minimizing.
I suppose I could strip jslint down.

What I really want is a command I can type inside a directory that
checks if the javascript files in that directory are minimizable and if
so minimizes them. Following that step all the files are gzipped. This
will save Apache the trouble of having to use mod_deflate for every
single request for a file. I don't know why they didn't build caching
for this into mod_deflate but since I have to do the minimize step
anyway I can also piggy back the compression step on the same command.
I know one thing for sure: I don't want to have to do all this
deployment stuff by hand starting with http://jslint.com anymore.

Peter

[Duncan Booth]

"Peter Michaux" <pe**********@g mail.comwrote:

It's more the comments, blank lines and beginning or mid-line spaces
that are a bigger concern. I have been using jslint to check for
problems before jsmin but jslint has many programmer preferences built
in and not just outright problems that will be harmful when minimizing.
I suppose I could strip jslint down.

I did a modified version of jslint which might interest you. You can find
it at http://codespeak.net/svn/kupu/trunk/kupu

The difference is that with my version there are command line options to
suppress most of the warnings so if you decide that you don't mind spurious
semicolons just use --ignore to suppress the ones you don't want. All the
warnings are numbered. See jslint.opts for the options I use for kupu, also
lint.py for a script which runs jslint.js over modified .js files only.
Also error messages all begin with filename and line number so they are
easily parsed by most editors.

Usage:
--
--browser, --nobrowser*
true if the standard browser globals should be predefined
--cap, --nocap*
true if upper case HTML should be allowed
--debug, --nodebug*
true if debugger statements should be allowed
--eqeqeq, --noeqeqeq*
true if === should be required
--error #
Specified messages are fatals
--evil, --noevil*
true if eval should be allowed
--extern #
Add external names
--help
show usage text
--ignore #
Specified messages are ignored
--jscript, --nojscript*
true if jscript deviations should be allowed
--laxLineEnd, --nolaxLineEnd*
true if line breaks should not be checked
--options #
Read additional arguments and options from a file
--passfail, --nopassfail*
true if the scan should stop on first error
--plusplus, --noplusplus*
true if increment/decrement should not be allowed
--redef, --noredef*
true if var redefinition should be allowed
--undef, --noundef*
true if undefined variables are errors
--warn #
Specified messages are warnings
--widget, --nowidget*
true if the Yahoo Widgets globals should be predefined

[VK]

Peter Michaux wrote:

Hi,

I'm thinking about code minimization. I can think of a few places where
whitespace matters

a + ++b
a++ + b
a - --b
a-- -b
when a line ends without a semi-colon in which case the new line
character matters.

Any others?

I guess Books of ECMA, Book of Tokens is the first place to look for
(ECMAScript 3rd ed., sec.7, "Lexical Conventions")

Besides that there is one cross-browser exception to add, but the book
above first.

[Peter Michaux]

Duncan Booth wrote:

"Peter Michaux" <pe**********@g mail.comwrote:

It's more the comments, blank lines and beginning or mid-line spaces
that are a bigger concern. I have been using jslint to check for
problems before jsmin but jslint has many programmer preferences built
in and not just outright problems that will be harmful when minimizing.
I suppose I could strip jslint down.

I did a modified version of jslint which might interest you. You can find
it at http://codespeak.net/svn/kupu/trunk/kupu

Thanks for the link. Your version may come in very handy.

Peter

[Peter Michaux]

VK wrote:

Peter Michaux wrote:

Hi,

I'm thinking about code minimization. I can think of a few places where
whitespace matters

a + ++b
a++ + b
a - --b
a-- -b
when a line ends without a semi-colon in which case the new line
character matters.

Any others?

I guess Books of ECMA, Book of Tokens is the first place to look for
(ECMAScript 3rd ed., sec.7, "Lexical Conventions")

Besides that there is one cross-browser exception to add, but the book
above first.

What is the exception?

Peter

[VK]

Peter Michaux wrote:

I guess Books of ECMA, Book of Tokens is the first place to look for
(ECMAScript 3rd ed., sec.7, "Lexical Conventions")

Besides that there is one cross-browser exception to add, but the book
above first.

What is the exception?

You are free to disregard, because it is not officially required by
specs, just an exploit of the internal tokenizer mechanics. Yet if it's
for public use - they you may account it as well. Any way, for long
string literals (too long to place on one line) instead of
concatenation one uses sometimes backslash trick:

var longString = "aaaaaaaaaaaaaa aaaaaa\
bbbbbbbbbbbbbbb bbbbbbbbbbbbbbb bbbb\
ccccccccccccccc ccccccccccccccc cccc"

If backslash is the very last character before the line break than the
tokenizer will be happy.

Saves a hell of cycles (no runtime concatenation) - but "hacky" of
course.

[Richard Cornford]

Peter Michaux wrote:

VK wrote:

>Peter Michaux wrote:

>>Hi,

I'm thinking about code minimization. I can think of a
few places where whitespace matters

a + ++b
a++ + b
a - --b
a-- -b
when a line ends without a semi-colon in which case the new
line character matters.

Any others?

I guess Books of ECMA, Book of Tokens is the first place to
look for (ECMAScript 3rd ed., sec.7, "Lexical Conventions")

Besides that there is one cross-browser exception to add,
but the book above first.

What is the exception?

Haven't you understood yet that VK has no understanding of javascript
and so does not know what is supposed to happen and what is not?
Whatever answer he gives you (if any) it is either going to be one of
his misunderstandin gs, or it will be total irrelevant to the subject of
determining which whitespace can safely be removed from javascript
source code.

It must also be worth pointing out that the more you encourage VK to
post (which is what you have just done) the more he will act to squander
the resources of the group in correcting his nonsense. Which will have
the consequence of denying you the time of the people who could
otherwise maybe usefully respond to your questions.

Richard.