Tree implementation, ideas for keyboard input method needed

B

Bubba

Greetings,

nodes of my ordered tree are implemented like this:

typedef struct __celltype {
labeltype label;
struct __celltype *first_child;
struct __celltype *next_sib;
struct __celltype *last_child;
int zadnji;
} celltype;
typedef celltype *node;
typedef celltype *TREE;

where first_child points to first sibling of the node, next_sib points to
next sibling and last_child points to last sibling in the node. Data
zadnji is true if the node is last child of any other node.

Labeltype is char.

Here's the whole implementation:

#include <stdlib.h>
#include <stdio.h>
#define LAMBDA NULL
#define labeltype char

typedef struct __celltype {
labeltype label;
struct __celltype *first_child;
struct __celltype *next_sib;
struct __celltype *last_child;
int zadnji; /*last*/
} celltype;

typedef celltype *node;
typedef celltype *TREE;

node MAKE_ROOT(labeltype l, TREE *T) {
if (!(*T = (celltype*) malloc(sizeof(celltype)))) {
printf("Nema dovoljno memorije.");
exit(1);
}
(*T)->label=l;
(*T)->first_child=NULL;
(*T)->next_sib=NULL;
(*T)->last_child=NULL;
(*T)->zadnji=0;
return *T;
}

node INSERT_CHILD(labeltype l, node i, TREE *T){
node temp;
if (!i) {
printf("Nema tog cvora.");
exit(1);
}
if (!(i->first_child= (celltype*) malloc(sizeof(celltype)))) {
printf("Nema dovoljno memorije.");
exit(1);
}
if (i->first_child==NULL) {
i->first_child->label=l;
i->first_child->first_child=NULL;
i->first_child->next_sib=NULL;
i->first_child->last_child=NULL;
i->first_child->zadnji=1;
}
if (i->first_child!=NULL) {
temp=i->first_child;
i->first_child->label=l;
i->first_child->first_child=NULL;
i->first_child->next_sib=temp;
i->first_child->last_child=NULL;
i->first_child->zadnji=0;
}
return i->first_child;
}

node ROOT(TREE T){
return T;
}

node INSERT_SIBLING(labeltype l, node i, TREE *T){
node temp;
if (!i) {
printf("Nema tog cvora.");
exit(1);
}
if (!(i->next_sib= (celltype*) malloc(sizeof(celltype)))) {
printf("Nema dovoljno memorije.");
exit(1);
}
if (i==ROOT(*T)) {
printf("Taj cvor je korijen.");
exit(1);
}
if(i->zadnji==1){
i->zadnji=0;
i->next_sib->label=l;
i->next_sib->first_child=NULL;
i->next_sib->next_sib=NULL;
i->next_sib->last_child=NULL;
i->next_sib->zadnji=1;
}
if(i->zadnji==0){
temp=i->next_sib;
i->next_sib->label=l;
i->next_sib->first_child=NULL;
i->next_sib->next_sib=temp;
i->next_sib->last_child=NULL;
i->next_sib->zadnji=0;
}
return i->next_sib;
}

node PARENT(node i,TREE T){
node c, d;
if (i==ROOT(T)) return LAMBDA;
if (!i) {
printf("Nema tog cvora.");
exit(1);
}
for (c=T->first_child; c!=NULL; c=c->next_sib)
if (c==i) return T;
for (c=T->first_child; c!=NULL; c=c->next_sib){
d=PARENT(i,c);
if (d!=NULL) return d;
}
return NULL;
}

void DELETE(node i, TREE *T) {
node c,d,e,f,temp;
if (i==ROOT(*T)) {
printf("Taj cvor je korijen."); /*is root*/
exit(1);
}
if (!i) {
printf("Nema tog cvora."); /*no node*/
exit(1);
}
if (i->first_child!=NULL) {
printf("Taj cvor ima djece."); /*has sib*/
exit(1);
}
c=PARENT(i,*T);
if (c->first_child==c->last_child) {
free(i);
c->first_child=NULL;
c->last_child=NULL;
}
if (i==c->first_child) {
temp=i->next_sib;
free(i);
c->first_child=temp;
}
if(i==c->last_child) {
d=c->first_child;
while(d!=c->last_child) {
if (d->next_sib==i) {
d->next_sib=NULL;
d->zadnji=1;
break;
}
d=d->next_sib;
}
free(i);
c->last_child=d;
}
if ((i!=c->first_child) && (i!=c->last_child)){

for (d=c->first_child; d!=NULL; d=d->next_sib){
if(d->next_sib==i) e=d;
if(i->next_sib==d) f=d;
}
free(i);
e->next_sib=f;
}
}

node FIRST_CHILD(node i, TREE T){
if (i->first_child==NULL) return NULL;
if (!i) {
printf("Nema tog cvora.");
exit(1);
}
return i->first_child;
}

node NEXT_SIBLING(node i, TREE T){
if (i->next_sib==NULL) return NULL;
if (!i) {
printf("Nema tog cvora.");
exit(1);
}
return i->next_sib;
}

labeltype LABEL(node i, TREE T) {
if (!i) {
printf("Nema tog cvora.");
exit(1);
}
return i->label;
}

void CHANGE_LABEL(labeltype l, node i, TREE *T){
if (!i) {
printf("Nema tog cvora.");
exit(1);
}
i->label=l;
}

void POSTORDER(node i, TREE T) {
node c;
for(c=FIRST_CHILD(i,T); c!=LAMBDA; c=NEXT_SIBLING(c,T))
POSTORDER(c,T);
printf("%c ", LABEL(i,T));
}


int main(void)
{
return 0;
}



Now I need a sane way to input a tree from keyboard. What would be the
easiest one?

Thanks in advance!
 
K

Keith Thompson

Just so you know, I'm not going to address your actual question, at
least for now. I do have some comments on your coding style.
nodes of my ordered tree are implemented like this:

typedef struct __celltype {
labeltype label;
struct __celltype *first_child;
struct __celltype *next_sib;
struct __celltype *last_child;
int zadnji;
} celltype;

Identifiers starting with underscores are reserved to the implementation
in most contexts. Identifiers starting with two underscores are
reserved in all contexts. You should never define such identifiers
yourself (unless you're writing code for a C compiler or library
implementation).

It's not necessary to use a typedef for a struct type. You could just
declare it as:

struct celltype {
...
};

and simply refer to it as "struct celltype". If you feel the need to
have a one-word name for the type, there's no need to use different
identifiers for the tag and typedef:

typedef struct celltype {
...
} celltype;
typedef celltype *node;
typedef celltype *TREE;

Typedefs for pointer types are generally a bad idea. The problem is
that they hide the fact that the type is a pointer. That's ok if it's a
truly opaque type, meaning that client code will not use any pointer
operations on it.

And logically, a pointer to a "celltype" isn't a node; it points to a node.

[...]
node MAKE_ROOT(labeltype l, TREE *T) {

All-caps names are conventionally used for macros. It's better style
to call the function "make_root".

And in fact I'd declare this as:

struct celltype *make_root(labeltype label, struct celltype **t);

Note that "l" is a poor identifier; it's too easily confused with the
digit "1".
if (!(*T = (celltype*) malloc(sizeof(celltype)))) {

The cast is unnecessary, and could be harmful (in certain circumstances,
it can inhibit important warnings). malloc() returns a void* result,
which can be implicitly converted to any pointer type (other than a
pointer-to-function type).

A good idiom is:

if ((*T = malloc(sizeof **T) != NULL) {
printf("Nema dovoljno memorije.");

Error mesages should be printed to stderr and terminated with "\n".

exit(1) doesn't necessarily indicate failure. Use exit(EXIT_FAILURE)
instead.

[...]
 
B

Ben Bacarisse

Bubba said:
Greetings,

nodes of my ordered tree are implemented like this:

typedef struct __celltype {

__celltype is a reserved name. celltype__ would be fine but then so
would celltype.
labeltype label;
struct __celltype *first_child;
struct __celltype *next_sib;
struct __celltype *last_child;
int zadnji;
} celltype;
typedef celltype *node;
typedef celltype *TREE;

where first_child points to first sibling of the node, next_sib points to
next sibling and last_child points to last sibling in the node. Data
zadnji is true if the node is last child of any other node.

Labeltype is char.

Here's the whole implementation:

#include <stdlib.h>
#include <stdio.h>
#define LAMBDA NULL
#define labeltype char

typedef struct __celltype {
labeltype label;
struct __celltype *first_child;
struct __celltype *next_sib;
struct __celltype *last_child;
int zadnji; /*last*/
} celltype;

typedef celltype *node;
typedef celltype *TREE;

node MAKE_ROOT(labeltype l, TREE *T) {
if (!(*T = (celltype*) malloc(sizeof(celltype)))) {

What's the cast for? I'd write:

if ((*T = malloc(sizeof **T)) == NULL)

That way you know, at first glance, that the size is correct.
printf("Nema dovoljno memorije.");
exit(1);
}

This code (allocate or fail with a message) is repeated lots of times.
Is a clear candidate for being a function.
(*T)->label=l;
(*T)->first_child=NULL;
(*T)->next_sib=NULL;
(*T)->last_child=NULL;
(*T)->zadnji=0;
return *T;
}

node INSERT_CHILD(labeltype l, node i, TREE *T){

You are using an uncommon naming convention (uppercase function names)
and unusual layout. Is there are a reason for this?
node temp;
if (!i) {
printf("Nema tog cvora.");
exit(1);
}
if (!(i->first_child= (celltype*) malloc(sizeof(celltype)))) {
printf("Nema dovoljno memorije.");
exit(1);
}
if (i->first_child==NULL) {
i->first_child->label=l;

This is obviously wrong. If i->first_child == NULL, you can't talk
about i->first_child->label.
i->first_child->first_child=NULL;
i->first_child->next_sib=NULL;
i->first_child->last_child=NULL;
i->first_child->zadnji=1;
}
if (i->first_child!=NULL) {
"else"?

temp=i->first_child;
i->first_child->label=l;
i->first_child->first_child=NULL;
i->first_child->next_sib=temp;
i->first_child->last_child=NULL;
i->first_child->zadnji=0;
}
return i->first_child;
}

The lack of spaces and the confusion caused by using tabs is making this
hard to read so I'll stop here.

Now I need a sane way to input a tree from keyboard. What would be the
easiest one?

I'd use parentheses:

(a b c)
(a (x (u v) z))

If you need to indicate "next_sib", use some
marker character:

(a b* c)
 
B

Bubba

Keith Thompson's log on stardate 16 sij 2012

I hoped for something like this in the first place, so thank you in
advance!
Identifiers starting with underscores are reserved to the
implementation in most contexts. Identifiers starting with two
underscores are reserved in all contexts. You should never define
such identifiers yourself (unless you're writing code for a C
compiler or library implementation).

What are the repercussions of choosing wrong number of underscores? :)
All-caps names are conventionally used for macros. It's better style
to call the function "make_root".

I'm aware of that, thanks, it will be corrected.
And in fact I'd declare this as:
struct celltype *make_root(labeltype label, struct celltype **t);

Could you please clarify the need for double pointer?
Note that "l" is a poor identifier; it's too easily confused with the
digit "1".

True again!
The cast is unnecessary, and could be harmful (in certain
circumstances, it can inhibit important warnings). malloc() returns
a void* result, which can be implicitly converted to any pointer type
(other than a pointer-to-function type).

I read some posts here about malloc's (non)casting. I'll do as you
suggested.
A good idiom is:
if ((*T = malloc(sizeof **T) != NULL) {
Ok.

Error mesages should be printed to stderr and terminated with "\n".
exit(1) doesn't necessarily indicate failure. Use exit(EXIT_FAILURE)
instead.

Thanks, I'll apply those too!
 
B

Bubba

Ben Bacarisse's log on stardate 16 sij 2012
The lack of spaces and the confusion caused by using tabs is making this
hard to read so I'll stop here.

Wow, this went wrong - I agree; I'll procede as you Keith suggested and
report back with sorted out code!

Thanks!
 
K

Kaz Kylheku

#define LAMBDA NULL

While this is a valiant effort, I recommend nothing less than:

#define PASTE(X, Y) X ## Y
#define XPASTE(X, Y) PASTE(X, Y)
#define MKTOKEN(A, B, C, D) XPASTE(XPASTE(A, B), XPASTE(C, D))
#define LAMBDA MKTOKEN(N, U, L, L)

Now you've REALLY defined a useless synonym for a useless synonym for zero,
in style!

Also, don't foget:

/* right-associatively-pasted lambda */
#define R_MKTOKEN(A, B, C, D) XPASTE(A, XPASTE(B, XPASTE(C, D)))
#define R_LAMBDA R_MKTOKEN(N, U, L, L)

/* left-associatively-pasted lambda */
#define L_MKTOKEN(A, B, C, D) XPASTE(XPASTE(XPASTE(A, B), C), D)
#define L_LAMBDA L_MKTOKEN(N, U, L, L)

Benchmark all three to see which LAMBDA null pointer constant runs faster!
 
K

Kaz Kylheku

Keith Thompson's log on stardate 16 sij 2012

I hoped for something like this in the first place, so thank you in
advance!


What are the repercussions of choosing wrong number of underscores? :)

The repercussions are that any of your compiler or library headers are
completely free to have things like this in them:

#define __celltype blorch[

This namespace does not belong to you, but to the implementors.

(Since C doesn't have namespaces, we have to draw these borders within the
identifiers themselves.)

Even if you haven no clash now with __celltype, you could have one tomorrow,
when you include some header that you did not previously include, port your
program to a different system, or upgrade (or downgrade) to a different version
of the compiler or any of your libraries, etc.
 
E

Eric Sosman

Greetings,

nodes of my ordered tree are implemented like this:

typedef struct __celltype {

A minor point: `__celltype' is a reserved identifier, and if
the implementation decides to use the identifier for its own
purposes (which it's at liberty to do without warning you), chaos
is likely to ensue. Use an identifier from the name space belonging
to the programmer -- `celltype' is fine, yes, even though you're also
going to use `celltype' as a typedef alias for `struct celltype'.
labeltype label;
struct __celltype *first_child;
struct __celltype *next_sib;
struct __celltype *last_child;
int zadnji;
} celltype;
typedef celltype *node;
typedef celltype *TREE;

Another minor point: It's been my experience that typedef aliases
for pointer types usually cause confusion. "Usually" is not "always,"
but I think you'd be better off without these aliases. The fact that
you have two identical aliases suggests that a certain amount of
confusion may already have crept in: When would you use `node' and
when would you use `TREE', and what difference are you trying to
express by choosing one or the other?
where first_child points to first sibling of the node, next_sib points to
next sibling and last_child points to last sibling in the node.

I'll assume you mean that `first_child' and `last_child' point
to the first and last *children* of the node, not to its first and
last siblings.
next sibling and last_child points to last sibling in the node. Data
zadnji is true if the node is last child of any other node.

Labeltype is char.

Here's the whole implementation:

#include<stdlib.h>
#include<stdio.h>
#define LAMBDA NULL
#define labeltype char

typedef struct __celltype {
labeltype label;
struct __celltype *first_child;
struct __celltype *next_sib;
struct __celltype *last_child;
int zadnji; /*last*/
} celltype;

typedef celltype *node;
typedef celltype *TREE;

node MAKE_ROOT(labeltype l, TREE *T) {
if (!(*T = (celltype*) malloc(sizeof(celltype)))) {

Yet more minor points: First, it's not necessary to cast the
result of malloc(). Second, the code relies on `TREE' being an
alias for `celltype*', which means you're not actually getting any
value out of the `TREE' alias. Here's a better paradigm, no matter
what type of data `T' points to:

if (!(*T = malloc(sizeof(**T))) ...
printf("Nema dovoljno memorije.");
exit(1);

Minor points: Put a newline '\n' at the end of the message,
and use exit(EXIT_FAILURE) instead of exit(1).

As a matter of design, it is usually not a good idea for a
"low-level" function to make such a drastic decision about the
fate of the entire program. Normally, it's better to indicate
failure in some other way, and let the higher-level caller decide
how to deal with it; the higher-level caller has more "knowledge"
of the state of the entire program. Even if the eventual action
is to exit(), the caller might want to do something else first.
Observe that malloc() does not terminate your program when it
fails, but instead reports the failure and lets the program make
its own choice; consider following that pattern in your own code,
especially in functions near the "leaves."
}
(*T)->label=l;
(*T)->first_child=NULL;
(*T)->next_sib=NULL;
(*T)->last_child=NULL;
(*T)->zadnji=0;
return *T;

Yet another confusion of aliases: This line relies on
`TREE' being the same as `node'. In one function you've used
three different names for exactly the same type! Do you think
this makes the code easier or harder to read?

There's also a design oddity. The function returns its
result via two different channels: It stores the new pointer
at the location pointed to by `T', and it also returns the
same pointer as the function value. There's nothing actually
"wrong" with this, but it's unusual enough to raise an eyebrow.
Usually, you'll want to use just one channel to export a result;
functions like time() are the exception, not the rule.
}

node INSERT_CHILD(labeltype l, node i, TREE *T){
node temp;
if (!i) {
printf("Nema tog cvora.");
exit(1);
}
if (!(i->first_child= (celltype*) malloc(sizeof(celltype)))) {
printf("Nema dovoljno memorije.");
exit(1);
}
if (i->first_child==NULL) {

Okay, here's an actual problem: Observe that (1) the code in
this part of the `if' statement will never execute, and (2) if it
somehow did execute you'd be trying to store data where the NULL
pointer points!
i->first_child->label=l;
i->first_child->first_child=NULL;
i->first_child->next_sib=NULL;
i->first_child->last_child=NULL;
i->first_child->zadnji=1;
}
if (i->first_child!=NULL) {

... and here's another part of the same problem: This code
will always execute, and will wind up producing an endless loop
in your data structure: That is, `i->first_child->next_sib' will
point back to `i->first_child' again. Any siblings that might
already have existed are now lost forever.
temp=i->first_child;
i->first_child->label=l;
i->first_child->first_child=NULL;
i->first_child->next_sib=temp;
i->first_child->last_child=NULL;
i->first_child->zadnji=0;
}
return i->first_child;
}

node ROOT(TREE T){
return T;
}

node INSERT_SIBLING(labeltype l, node i, TREE *T){
node temp;
if (!i) {
printf("Nema tog cvora.");
exit(1);
}
if (!(i->next_sib= (celltype*) malloc(sizeof(celltype)))) {
printf("Nema dovoljno memorije.");
exit(1);
}
if (i==ROOT(*T)) {
printf("Taj cvor je korijen.");
exit(1);
}
if(i->zadnji==1){
i->zadnji=0;
i->next_sib->label=l;
i->next_sib->first_child=NULL;
i->next_sib->next_sib=NULL;
i->next_sib->last_child=NULL;
i->next_sib->zadnji=1;
}
if(i->zadnji==0){
temp=i->next_sib;
i->next_sib->label=l;
i->next_sib->first_child=NULL;
i->next_sib->next_sib=temp;
i->next_sib->last_child=NULL;
i->next_sib->zadnji=0;
}
return i->next_sib;
}

node PARENT(node i,TREE T){

If you expect to navigate from child to parent, wouldn't it
be easier to store that link directly in the node than to do a
recursive search for it?
node c, d;
if (i==ROOT(T)) return LAMBDA;
if (!i) {
printf("Nema tog cvora.");
exit(1);
}
for (c=T->first_child; c!=NULL; c=c->next_sib)
if (c==i) return T;
for (c=T->first_child; c!=NULL; c=c->next_sib){
d=PARENT(i,c);
if (d!=NULL) return d;
}
return NULL;
}

I'm running out of energy, and won't say much more; absence
of commentary should not be construed as approval! One thing that
stands out is that you've duplicated the node-building operation
(allocate memory and initialize the struct) in about five places;
it would be much simpler (and easier!) to write just one function
to do this job, and to call it from wherever it's needed. You've
got three functions whose job is to create a new struct and attach
it in various ways; consider separating the "create" and "attach"
operations, and I think you'll have shorter and cleaner code.
void DELETE(node i, TREE *T) {
node c,d,e,f,temp;
if (i==ROOT(*T)) {
printf("Taj cvor je korijen."); /*is root*/
exit(1);
}
if (!i) {
printf("Nema tog cvora."); /*no node*/
exit(1);
}
if (i->first_child!=NULL) {
printf("Taj cvor ima djece."); /*has sib*/
exit(1);
}
c=PARENT(i,*T);
if (c->first_child==c->last_child) {
free(i);
c->first_child=NULL;
c->last_child=NULL;
}
if (i==c->first_child) {
temp=i->next_sib;
free(i);
c->first_child=temp;
}
if(i==c->last_child) {
d=c->first_child;
while(d!=c->last_child) {
if (d->next_sib==i) {
d->next_sib=NULL;
d->zadnji=1;
break;
}
d=d->next_sib;
}
free(i);
c->last_child=d;
}
if ((i!=c->first_child)&& (i!=c->last_child)){

for (d=c->first_child; d!=NULL; d=d->next_sib){
if(d->next_sib==i) e=d;
if(i->next_sib==d) f=d;
}
free(i);
e->next_sib=f;
}
}

node FIRST_CHILD(node i, TREE T){
if (i->first_child==NULL) return NULL;
if (!i) {
printf("Nema tog cvora.");
exit(1);
}
return i->first_child;
}

node NEXT_SIBLING(node i, TREE T){
if (i->next_sib==NULL) return NULL;
if (!i) {
printf("Nema tog cvora.");
exit(1);
}
return i->next_sib;
}

labeltype LABEL(node i, TREE T) {
if (!i) {
printf("Nema tog cvora.");
exit(1);
}
return i->label;
}

void CHANGE_LABEL(labeltype l, node i, TREE *T){
if (!i) {
printf("Nema tog cvora.");
exit(1);
}
i->label=l;
}

void POSTORDER(node i, TREE T) {
node c;
for(c=FIRST_CHILD(i,T); c!=LAMBDA; c=NEXT_SIBLING(c,T))
POSTORDER(c,T);
printf("%c ", LABEL(i,T));
}


int main(void)
{
return 0;
}



Now I need a sane way to input a tree from keyboard. What would be the
easiest one?

You could use parentheses to enclose groups of siblings:

(A(B,C),D,E(F(G,H),I),J)

could encode a tree that might also be diagrammed as

A
+- B
+- C
D
E
+- F
+- G
+- H
+- I
J

But first, fix your code!
 
B

Ben Bacarisse

Keith Thompson said:
A good idiom is:

if ((*T = malloc(sizeof **T) != NULL) {

There's a missing ')' of course:

if ((*T = malloc(sizeof **T)) != NULL) {

<snip>
 
J

James Kuyper

Keith Thompson's log on stardate 16 sij 2012

I hoped for something like this in the first place, so thank you in
advance!


What are the repercussions of choosing wrong number of underscores? :)

Theoretically: the behavior of your program is undefined. This means
that the C standard places no restrictions on the behavior of the
resulting program. That's a bad thing, if there's any particular
behavior you want your program to have, since it might not have that
behavior. It's an even worse thing if there are any things your program
could do that you don't want it to do (such as erasing all of your
files), because there's a chance that it might do one of those things.

Practically: identifiers are reserved to the implementation, so they can
by used by the implementation for it's own purpose, without worrying
about the possibility that user code might interfere. Consider what
could go wrong with your program if __celltype is defined by the
implementation, inside stdlib.h, for some purpose. The best possible
outcome is that the compiler will notice two different conflicting
definitions for __celltype, and complain. The worst possibility is that
__celltype is defined in some way that produces no warning messages when
you misuse it.
Could you please clarify the need for double pointer?

Please don't refer to that as a "double pointer". It's a pointer to a
pointer. a "double pointer" could be interpreted as referring to a
"double*".

I'll let Keith answer the rest of your question in more detail, since
it's his suggestion.
 
B

Ben Bacarisse

William Ahern said:
You mean confusion caused by *not* using tabs, or by software trying to
convert perfectly fine tabs to spaces and--not surprisingly--fscking the
indentation?

By "using" I meant "posting with". Tabs in a post are unlikely to look
the same as tabs in a code editor. I made the remark when I was looking
at some code that had tabs, though I see there are some lines that looks
odd and have no tabs (in the post).
Tabs for indentation, spaces for alignment. Almost every editor gets this
right by default. It's takes a concerted effort to screw it up. (I'm not
laying blame on anyone in particular here. I'm presuming it's the OP's
client software that is causing the problem.)

We are probably in agreement. I don't care what the editor uses for the
code, but once tabs get into a post, you've almost always lost.
 
K

Keith Thompson

William Ahern said:
You mean confusion caused by *not* using tabs, or by software trying to
convert perfectly fine tabs to spaces and--not surprisingly--fscking the
indentation?

No, I'm fairly sure he means confusion caused by using tabs.
Tabs for indentation, spaces for alignment. Almost every editor gets this
right by default. It's takes a concerted effort to screw it up. (I'm not
laying blame on anyone in particular here. I'm presuming it's the OP's
client software that is causing the problem.)

I've worked on far too much code that has a random mixture of tabs
and spaces, with a tab representing a variable number of columns,
depending on who modified a particular line of code most recently.
My personal solution to the problem is to use spaces exclusively
for indentation -- and that was part of the coding standard at my
most recent job. (If I worked in environment that required the
exclusive use of tabs for indentation, of course I'd use tabs.)

I'm not going to argue that you should follow my convention, but
you should be aware that there are perfectly legitimate reasons
for using spaces, even if you don't agree with them; people who use
spaces rather than tabs aren't necessarily doing so out of ignorance.

And tabs can cause serious formatting problems when posting to
Usenet, because there's news software that doesn't handle them
correctly.
 
K

Keith Thompson

Bubba said:
Keith Thompson's log on stardate 16 sij 2012
I hoped for something like this in the first place, so thank you in
advance!


What are the repercussions of choosing wrong number of underscores? :)

Undefined behavior.

If you're *lucky*, you'll happen to collide with an
implementation-defined identifier and your code will fail to compile.

If you're *unlucky*, your program works as you intended it to work, or
it will fail in subtle ways that are difficult to find.

The compiler and/or standard runtime library can do anything they like
with identifiers that start with "__", and nearly anything they like
with identifiers that start with a single underscore. If you define
such an identifier yourself, it may or may not conflict with a
system-defined identifier.

So don't do that.

[...]
Could you please clarify the need for double pointer?

A "double pointer" could be either a pointer to a pointer, or a pointer
to type "double". It's better to refer to it as a pointer-to-pointer.

Why a pointer to a pointer? Don't ask me, it's your code; you declared
it with a pointer-to-pointer yourself. All I did was replace the
typedef (which hides the fact that the type is a pointer) with its full
definition.

But in general, if you want a function to modify something of type foo,
you need to pass it a foo*. In this case, "foo" happens to be a pointer
type itself.

[...]
 
K

Keith Thompson

[...]
And if you have a friendly editor it will convert tabs to spaces and
vice versa just as you desire.

But not very well if different past maintainers had different ideas of
where the tabstops should be. For example, one section of code might
assume 2-column tabstops, and another might assume 4-column or 8-column
tabstops.

Which is ok if tabs (and only tabs) are used exclusively for
indentation, but not if both sections have a mix of tabs and spaces.
 
I

ImpalerCore

Greetings,
[snip]

Now I need a sane way to input a tree from keyboard. What would be the
easiest one?

Thanks in advance!

The simplest is probably to leverage an existing XML parser and write
your own XML format to populate your own custom "DOM" structure. The
expat and libxml2 are fairly popular, and there are some varying
implementations of DOM floating around.

If you want to input a tree from the keyboard, you could type in the
tree data in XML format, feed it to one of the parsers, and use its
events to trigger using your interface to populate your tree. It'll
take some work, but learning how to use an XML parser can be quite
useful.

Best regards,
John D.
 
K

Kaz Kylheku

And if you have a friendly editor it will convert tabs to spaces and
vice versa just as you desire.

I use program that I call autotab. This is integrated into
Vim. When I open a file, autotab is called to analyze up to 5000
lines of text from the file and emit a configuration for Vim,
such as whether tabs are expanded with spaces (expandtab), the soft indentation
size (shiftwidth) and the hard tab size (tabstop).

The code is here:

http://www.kylheku.com/cgit/c-snippets/tree/autotab.c

This program is so rarely wrong that I've stopped fixing the situations when it
is wrong.
 
E

Eric Sosman

Greetings,
[snip]

Now I need a sane way to input a tree from keyboard. What would be the
easiest one?

Thanks in advance!

The simplest is probably to leverage an existing XML parser and write
your own XML format to populate your own custom "DOM" structure. The
expat and libxml2 are fairly popular, and there are some varying
implementations of DOM floating around.

Angels and ministers of grace defend us! The O.P. is trying
to build a tree in which each node carries a payload of

ONE

LONESOME

`CHAR'

VALUE

.... and you want to inflict XML on him?!?!

Whereabouts do you live, ImpalerCore? I want to be about five
hundred miles away, and upwind, when you find a cockroach in your
cupboard and exterminate it with nuclear armaments.
 
K

Keith Thompson

[email protected] (Richard Harter) said:
[...]
And if you have a friendly editor it will convert tabs to spaces and
vice versa just as you desire.

But not very well if different past maintainers had different ideas of
where the tabstops should be. For example, one section of code might
assume 2-column tabstops, and another might assume 4-column or 8-column
tabstops.

Within a single file?
Yes.

And your editor displays them all correctly?

No.

That was the problem.
How remarkable.

I don't you caught what I had in mind. If your editor has the option
of automatically converting tabs into spaces you can use tabs with the
settings you prefer, get the indentation that floats your boat, and
still have a tab free file. I keep all of my code in tab free format
(except for makefiles - boo, hiss). If your editor can go the other
way, i.e., convert those pesky spaces into tabs, you can deal with
files that historically have tabs that you want to keep that way.
Just convert tabs to spaces, edit, and convert back from spaces to
tabs.

That could work if everyone who modifies the file *either* uses only
tabs for indentation *or* uses only spaces (with a consistent number
of columns per level) for indentation. My recent experience does
not indicate that maintaining that kind of consistency is practical.
Perhaps in some organizations it would be.
 

Ask a Question

Want to reply to this thread or ask your own question?

You'll need to choose a username for the site, which only take a couple of moments. After that, you can post your question and our members will help you out.

Ask a Question

Similar Threads

Beginner at c 0
A process take input from /proc/<pid>/fd/0, but won't process it 0
simplebinary tree 7
Fibonacci 0
Binary Search Tree implementation. 0
tree 6
Lexical Analysis on C++ 1
Balancing a Binary Tree 4

Members online

Forum statistics

Threads
473,744
Messages
2,569,483
Members
44,901
Latest member
Noble71S45

Latest Threads

Top