multiline comments

[Edward Elliott]

At the risk of flogging a dead horse, I'm wondering why Python doesn't have
any multiline comments. One can abuse triple-quotes for that purpose, but
that's obviously not what it's for and doesn't nest properly. ML has a
very elegant system for nested comments with (* and *).

Using an editor to throw #s in front of every line has limitations. Your
editor has to support it and you have to know how to use that feature. Not
exactly intuitive or easy for novices to pick up. Also a pain if your
preferred editor is python/perl/sh-agnostic.

Saying coders shouldn't use multiline comments to disable code misses the
point. Coders will comment out code regardless of the existence of
multiline comemnts. There has to be a better argument for leaving them out.

Keeping the language small and simple is desirable, but it's not an
absolute. A little syntactic sugar like 'for x in s' makes code easier to
read than 'for i in len(s): x = s'. So what are the tradeoffs involved
with nested multiline comments? I'd like to understand the reasoning
behind keeping them out.

 

[James Stroud]

At the risk of flogging a dead horse, I'm wondering why Python doesn't
have any multiline comments. One can abuse triple-quotes for that
purpose, but that's obviously not what it's for and doesn't nest
properly. ML has a very elegant system for nested comments with (* and *).

Using an editor to throw #s in front of every line has limitations.
Your editor has to support it and you have to know how to use that
feature. Not exactly intuitive or easy for novices to pick up. Also a
pain if your preferred editor is python/perl/sh-agnostic.

Saying coders shouldn't use multiline comments to disable code misses
the point. Coders will comment out code regardless of the existence of
multiline comemnts. There has to be a better argument for leaving them
out.

Keeping the language small and simple is desirable, but it's not an
absolute. A little syntactic sugar like 'for x in s' makes code easier
to read than 'for i in len(s): x = s'. So what are the tradeoffs
involved with nested multiline comments? I'd like to understand the
reasoning behind keeping them out.

I think the absence of multiline comments is like the requirement for
indentation. It enforces good habits. Better is to make your multiple
lines a function and comment out the function call.

James

--
James Stroud
UCLA-DOE Institute for Genomics and Proteomics
Box 951570
Los Angeles, CA 90095

http://www.jamesstroud.com/

 

[Ben Finney]

At the risk of flogging a dead horse, I'm wondering why Python
doesn't have any multiline comments. [...]

Using an editor to throw #s in front of every line has
limitations. Your editor has to support it and you have to know
how to use that feature. Not exactly intuitive or easy for
novices to pick up. Also a pain if your preferred editor is
python/perl/sh-agnostic.

I think the absence of multiline comments is like the requirement
for indentation. It enforces good habits. Better is to make your
multiple lines a function and comment out the function call.

And/or switch to an editor that can perform editing operations on a
range of lines.

 

[Atanas Banov]

At the risk of flogging a dead horse, I'm wondering why Python doesn't have
any multiline comments. One can abuse triple-quotes for that purpose, but
that's obviously not what it's for and doesn't nest properly. ....
Saying coders shouldn't use multiline comments to disable code misses the
point. Coders will comment out code regardless of the existence of
multiline comemnts. There has to be a better argument for leaving them out.

i beg to differ: you'd be surprised how much effect can little
inconveniences have.

want to comment block of code? use tripple-quotes. does not nest? ahhh,
maybe it's time to get rid of that block you commented out a month ago
"just in case the new code doesnt work".

that gives you incentive to tidy up. don't be a code slob... don't
leave a mess forever ;-)

 

[Ben Finney]

i beg to differ: you'd be surprised how much effect can little
inconveniences have.

want to comment block of code? use tripple-quotes. does not nest?
ahhh, maybe it's time to get rid of that block you commented out a
month ago "just in case the new code doesnt work".

Indeed. Using revision control means never needing to comment out
blocks of code.

If your revision control system is so inconvenient to use that you'd
rather have large blocks of commented-out code, it's time to start
using a better RCS -- perhaps a distributed one, so you can commit to
your own local repository with abandon while trying out changes.

 

[Edward Elliott]

want to comment block of code? use tripple-quotes. does not nest? ahhh,
maybe it's time to get rid of that block you commented out a month ago
"just in case the new code doesnt work".

that gives you incentive to tidy up. don't be a code slob... don't
leave a mess forever ;-)

And when the section I want to comment out contains a legit doc string in
the middle, triple-quotes won't work. There are valid reasons to nest
comments which have nothing to do with laziness or sloppy code.

Forcing programmers to write clean code with syntax is like teaching a pig
to sing: it wastes your time and annoys the pig. Good coding is a state of
mind, not a parser option.

 

[Edward Elliott]

And/or switch to an editor that can perform editing operations on a
range of lines.

I'm not unsympathetic to this point of view, as I would feel hamstrung
without my vim. It's more that I object to the paternalism of telling
people they have to use such an editor. There are times when notepad is
all that's available.

On top of that, the expressive power of nested comments seems greater than
an endless string of ^#s. Sometimes it's just easier to see what's going on.

 

[Edward Elliott]

Indeed. Using revision control means never needing to comment out
blocks of code.

Typing (* and *) on a few line will always be quicker, easier, and less
confusing than any rcs diffs/restores. Once you delete the code you can no
longer see it or add pieces back in without retrieving it from an external
store. I'm not saying nested comments solve every problem, just that there
exists a certain (perhaps small) class of problems they solve particularly
well.

Personally, I rarely leave code commented out beyond a single coding
session. But my particular coding habits aren't relevant here.

 

[Lawrence D'Oliveiro]

ML has a
very elegant system for nested comments with (* and *).

Which, if you mistype an opening or closing comment symbol, can lead to
some very mysterious syntax errors.

 

[Duncan Booth]

Typing (* and *) on a few line will always be quicker, easier, and
less confusing than any rcs diffs/restores. Once you delete the code
you can no longer see it or add pieces back in without retrieving it
from an external store. I'm not saying nested comments solve every
problem, just that there exists a certain (perhaps small) class of
problems they solve particularly well.

Would you care to name a few languages which support nested block
comments? There really aren't many: ML as you mentioned; Standard Pascal
doesn't permit nesting of comments but *some* implementations do allow it.

Want to comment out a block of code in C++? The only (nearly) reliable way
is to insert single-line comments down the block. You can't use a block
comment if there are any other block comments inside the code you want to
block out.

The danger of block comments is that if you forget to close the comment
you can accidentally comment out a large part of your code. With support
from the editor (coloured highlighting of comments) this isn't so bad,
but then if you have a decent editor you don't need the block comments
anyway as you will be able to comment/uncomment a block in your editor.

Doc strings will usually work as an alternative, especially since you
have a choice of two flavours of triple quoted strings, so if you use
one for docstrings the other is always free for your temporary block
comments.

Forcing programmers to write clean code with syntax is like teaching a
pig to sing: it wastes your time and annoys the pig.

This pig gets much more annoyed having to maintain code which has large
chunks of unneeded commented out code left over from some other programmer,
or which has completely messed up indentation.

 

[Roel Schroeven]

Duncan Booth schreef:

Would you care to name a few languages which support nested block
comments? There really aren't many: ML as you mentioned; Standard Pascal
doesn't permit nesting of comments but *some* implementations do allow it.

Want to comment out a block of code in C++? The only (nearly) reliable way
is to insert single-line comments down the block. You can't use a block
comment if there are any other block comments inside the code you want to
block out.

Depends, some compilers support that. But the preferred way, which works
very well, is to use preprocessor directives:

#if 0
...
#endif

Works in both C and C++.

 

[Edward Elliott]

Want to comment out a block of code in C++? The only (nearly) reliable way
is to insert single-line comments down the block. You can't use a block
comment if there are any other block comments inside the code you want to
block out.

As Roel said, #if 0 is the standard way. It abuses the preprocessor and
doesn't show up in syntax highlighting, but other than that works very
well. Honestly though, /* and */ should have nested properly since day 1.
Adding it wouldn't even break existing code.

The danger of block comments is that if you forget to close the comment
you can accidentally comment out a large part of your code.

No, unclosed comments should raise a syntax error. Would you accept an
unclosed string literal?

Doc strings will usually work as an alternative, especially since you
have a choice of two flavours of triple quoted strings, so if you use
one for docstrings the other is always free for your temporary block
comments.

That's a fair point, if a bit of a kludge. 90% there is good enough in
practice.

This pig gets much more annoyed having to maintain code which has large
chunks of unneeded commented out code left over from some other programmer,
or which has completely messed up indentation.

Sure they can be abused. So can a thousand other language features. My
point is you can't teach good coding through syntax, and trying to causes
more problems than it solves.

I would argue the current system is in fact slightly worse, because people
will comment out code chunks anyway (either lots of #s or triple-quotes)
and are less likely to remove them when it's more work. But either way,
social pressure is infinitely more effective at cleaning up code than
comment syntax.

 

[Gregor Horvath]

On top of that, the expressive power of nested comments seems greater
than an endless string of ^#s. Sometimes it's just easier to see what's
going on.

not if you are using grep

 

[Peter Tillotson]

Indeed. Using revision control means never needing to comment out
blocks of code.

If your revision control system is so inconvenient to use that you'd
rather have large blocks of commented-out code, it's time to start
using a better RCS -- perhaps a distributed one, so you can commit to
your own local repository with abandon while trying out changes.

I'm not sure I agree, revision control is great but not the only answer.
In multi-developer teams working on the trunk, it its kind of
inconvenient if someone checks in broken code. It also blocks critical
path development if the person responsible for the code you conflict
with happens to be out on holiday. Block commenting is a clear flag to
that developer that something has changed - ideally he'd notice, see
sensible revision control comments, see the flag on the wiki or you
would remember to tell him. But if all of that fails, if it is commented
in the code it should get picked up at a code review.

Personally, I prefer clear code, minimally commented with good high
level descriptions of particularly complex section / algorithms. The
later doesn't always fit neatly on one line. There is an argument that
these should go into their own functions and be commented at the
function level. Again I'm not sure I agree entirely - function comments
that are auto extracted to create api docs (sorry Java background )
need only outline What a function does. There is a place for multiline
comments to describe How that is achieved.

Having said all that, I generally don't like comments, they are often
maintained poorly, too numerous, too verbose (red rag ) - so i'm
undecided whether they should be made easier for developers or
discouraged except where vital. Perhaps we should make them really hard
and elegant - mandate latex/mathml markup so good editors can display
the equations we are implementing

 

[Jorge Godoy]

Typing (* and *) on a few line will always be quicker, easier, and less
confusing than any rcs diffs/restores. Once you delete the code you can
no longer see it or add pieces back in without retrieving it from an
external store.

Try using Subversion. You can work and make diffs disconnected from the
network. You can't, of course, commit / update but you can work with it
and have what you need to compare original code (i.e. the one from the last
commit) to new code and go back to original code if needed.

I'm not saying nested comments solve every problem, just that
there exists a certain (perhaps small) class of problems they solve
particularly well.

I don't miss them.

Personally, I rarely leave code commented out beyond a single coding
session. But my particular coding habits aren't relevant here.

Well, I believe they are since it looks like a habit of yours to use
multiline comments. It is common for people coming from other programming
languages that support them.

--
Jorge Godoy <[email protected]>

"Quidquid latine dictum sit, altum sonatur."
- Qualquer coisa dita em latim soa profundo.
- Anything said in Latin sounds smart.

 

[Jorge Godoy]

Sure they can be abused. So can a thousand other language features. My
point is you can't teach good coding through syntax, and trying to causes
more problems than it solves.

I like the phrase: there are some languages that incentivates bad practices
in programming; there is Python that doesn't.

Some rigid syntax indeed makes you think one way -- and that's one of
Python's motto, isn't it? "There's one right way to do it" -- but that
will make your code more understandable and readable in the future.

I would argue the current system is in fact slightly worse, because people
will comment out code chunks anyway (either lots of #s or triple-quotes)
and are less likely to remove them when it's more work. But either way,
social pressure is infinitely more effective at cleaning up code than
comment syntax.

Is it harder to remove "n" lines of code commented out with "#" than "n"
lines of multiline commented code? How? The same question goes for triple
quoted code.

--
Jorge Godoy <[email protected]>

"Quidquid latine dictum sit, altum sonatur."
- Qualquer coisa dita em latim soa profundo.
- Anything said in Latin sounds smart.

 

[Jorge Godoy]

I'm not sure I agree, revision control is great but not the only answer.
In multi-developer teams working on the trunk, it its kind of
inconvenient if someone checks in broken code. It also blocks critical

This is something that should be a policy: no untested and working code
should be commited to the trunk; if you need to commit it for any reason
create a branch and do it there.

path development if the person responsible for the code you conflict
with happens to be out on holiday. Block commenting is a clear flag to

Here I believe that no code revision system is a substitute for project. If
you have two conflicting changes on the same line of code, then you surely
are missing some discussion on the code and more documentation on what is
being done / has been done. Revision management is no substitute for
meetings and projects.

that developer that something has changed - ideally he'd notice, see
sensible revision control comments, see the flag on the wiki or you
would remember to tell him. But if all of that fails, if it is commented
in the code it should get picked up at a code review.

I believe that it is easier to see multiple commented out lines than just
the beginning and ending of a multiline comment. Specially when you're
screening the code instead of reading it line by line.

Personally, I prefer clear code, minimally commented with good high
level descriptions of particularly complex section / algorithms. The

We have the same taste, except that I prefer documenting more things than
just complex algorithms so I have a lot of comments and docstrings in my
code.

later doesn't always fit neatly on one line. There is an argument that
these should go into their own functions and be commented at the
function level. Again I'm not sure I agree entirely - function comments
that are auto extracted to create api docs (sorry Java background )
need only outline What a function does. There is a place for multiline
comments to describe How that is achieved.

I still believe that you're working with an inappropriate environment if
your editor can't use some extension for the language you choose (coming
from a Java background you might like PyDev on Eclipse, even though its
indentation features aren't as nice as Emacs' features...) or being able to
repeat the comment symbol from one line to the next when it wraps (keeping
indentation, of course!)

Having said all that, I generally don't like comments, they are often
maintained poorly, too numerous, too verbose (red rag ) - so i'm
undecided whether they should be made easier for developers or
discouraged except where vital. Perhaps we should make them really hard
and elegant - mandate latex/mathml markup so good editors can display
the equations we are implementing

There's an approach that allows using those... I don't remember which
docsystem allows for MathML markup. But then, I'd go with DocBook + MathML
+ SVG ;-) (Hey! You started! And you even said that you didn't like
verbose comments... ;-))

--
Jorge Godoy <[email protected]>

"Quidquid latine dictum sit, altum sonatur."
- Qualquer coisa dita em latim soa profundo.
- Anything said in Latin sounds smart.

 

[Jorge Godoy]

And when the section I want to comment out contains a legit doc string in
the middle, triple-quotes won't work. There are valid reasons to nest

You can use either """ or '''. I don't keep changing them in my code, so I
can always use the other type (usually I use " so for commenting things out
I'd use ') to do that.

comments which have nothing to do with laziness or sloppy code.

Forcing programmers to write clean code with syntax is like teaching a pig
to sing: it wastes your time and annoys the pig. Good coding is a state
of mind, not a parser option.

If the latter can help, why not?

--
Jorge Godoy <[email protected]>

"Quidquid latine dictum sit, altum sonatur."
- Qualquer coisa dita em latim soa profundo.
- Anything said in Latin sounds smart.

 

[Peter Tillotson]

nice one Jorge

This is something that should be a policy: no untested and working code
should be commited to the trunk; if you need to commit it for any reason
create a branch and do it there.

typo

committing broken code to trunk should punishable by stocks at least --
perhaps a public flogging.

Though on a serious note, you need a good reason for creating arbitrary
branches and a clearly defined naming policy. You also need to remember
to get off the branch asap before you inadvertently start a major fork.

Here I believe that no code revision system is a substitute for project. If
you have two conflicting changes on the same line of code, then you surely
are missing some discussion on the code and more documentation on what is
being done / has been done. Revision management is no substitute for
meetings and projects.


I believe that it is easier to see multiple commented out lines than just
the beginning and ending of a multiline comment. Specially when you're
screening the code instead of reading it line by line.


We have the same taste, except that I prefer documenting more things than
just complex algorithms so I have a lot of comments and docstrings in my
code.

I still believe that you're working with an inappropriate environment if
your editor can't use some extension for the language you choose (coming
from a Java background you might like PyDev on Eclipse, even though its
indentation features aren't as nice as Emacs' features...) or being able to
repeat the comment symbol from one line to the next when it wraps (keeping
indentation, of course!)

Sadly i don't get to code at the coalface much recently. I've tinkered
in python with pydev and vi over the last couple of years - i just
really dislike coding on a white background. I'm sure eclipse can do it
- i'm just not sure i've got the perseverance to work out how.

There's an approach that allows using those... I don't remember which
docsystem allows for MathML markup. But then, I'd go with DocBook + MathML
+ SVG ;-) (Hey! You started! And you even said that you didn't like
verbose comments... ;-))

oooh - but surely a picture is worth a thousand words and with SVG no
truer word was spoken

I was only half kidding about latex. I can just about read pure latex
but that human readable xml stuff always defeats me

 

[Sion Arrowsmith]

On top of that, the expressive power of nested comments seems greater than
an endless string of ^#s. Sometimes it's just easier to see what's going on.

Really? Under what circumstances is it easier to see what's going on
with start/end comments than with comment-to-end-of-line?