[PATCH] hash: Remove useless init_hash()

[PATCH] hash: Remove useless init_hash()

[Stephen Boyd]

init_hash() is essentially a memset() so just use that.

Signed-off-by: Stephen Boyd <bebarino@gmail.com>
---
 Documentation/technical/api-hash.txt |    4 ----
 diffcore-rename.c                    |    2 +-
 hash.h                               |    7 -------
 3 files changed, 1 insertions(+), 12 deletions(-)

diff --git a/Documentation/technical/api-hash.txt b/Documentation/technical/api-hash.txt
index e5061e0..7cf64ec 100644
--- a/Documentation/technical/api-hash.txt
+++ b/Documentation/technical/api-hash.txt
@@ -23,10 +23,6 @@ Data Structures
 Functions
 ---------
 
-`init_hash`::
-
-	Initialize the hash table.
-
 `free_hash`::
 
 	Release memory associated with the hash table.
diff --git a/diffcore-rename.c b/diffcore-rename.c
index df41be5..b355520 100644
--- a/diffcore-rename.c
+++ b/diffcore-rename.c
@@ -382,7 +382,7 @@ static int find_exact_renames(void)
 	int i;
 	struct hash_table file_table;
 
-	init_hash(&file_table);
+	memset(&file_table, 0, sizeof(file_table));
 	for (i = 0; i < rename_src_nr; i++)
 		insert_file_table(&file_table, -1, i, rename_src[i].one);
 
diff --git a/hash.h b/hash.h
index 69e33a4..418be24 100644
--- a/hash.h
+++ b/hash.h
@@ -33,11 +33,4 @@ extern void **insert_hash(unsigned int hash, void *ptr, struct hash_table *table
 extern int for_each_hash(const struct hash_table *table, int (*fn)(void *));
 extern void free_hash(struct hash_table *table);
 
-static inline void init_hash(struct hash_table *table)
-{
-	table->size = 0;
-	table->nr = 0;
-	table->array = NULL;
-}
-
 #endif
-- 
1.7.2.19.g9a302

Re: [PATCH] hash: Remove useless init_hash()

[Jakub Narebski]

Stephen Boyd <bebarino@gmail.com> writes:

> init_hash() is essentially a memset() so just use that.
> 
> Signed-off-by: Stephen Boyd <bebarino@gmail.com>
[...]

Encapsulation is good.

> diff --git a/diffcore-rename.c b/diffcore-rename.c
> index df41be5..b355520 100644
> --- a/diffcore-rename.c
> +++ b/diffcore-rename.c
> @@ -382,7 +382,7 @@ static int find_exact_renames(void)
>  	int i;
>  	struct hash_table file_table;
>  
> -	init_hash(&file_table);
> +	memset(&file_table, 0, sizeof(file_table));

This is IMHO slightly less readable, and doesn't protect against
changes in implementation.

>  	for (i = 0; i < rename_src_nr; i++)
>  		insert_file_table(&file_table, -1, i, rename_src[i].one);
>  
> diff --git a/hash.h b/hash.h
> index 69e33a4..418be24 100644
> --- a/hash.h
> +++ b/hash.h
> @@ -33,11 +33,4 @@ extern void **insert_hash(unsigned int hash, void *ptr, struct hash_table *table
>  extern int for_each_hash(const struct hash_table *table, int (*fn)(void *));
>  extern void free_hash(struct hash_table *table);
>  
> -static inline void init_hash(struct hash_table *table)
> -{
> -	table->size = 0;
> -	table->nr = 0;
> -	table->array = NULL;
> -}

*This* could be replaced by memset.

> -
>  #endif
> -- 
> 1.7.2.19.g9a302
> 

-- 
Jakub Narebski
Poland
ShadeHawk on #git

Re: [PATCH] hash: Remove useless init_hash()

[Ævar Arnfjörð Bjarmason]

On Tue, Jul 27, 2010 at 10:30, Jakub Narebski <jnareb@gmail.com> wrote:

CC-ing Linus since he wrote it (per Documentation/SubmittingPatches).

> Stephen Boyd <bebarino@gmail.com> writes:
>
>> init_hash() is essentially a memset() so just use that.
>>
>> Signed-off-by: Stephen Boyd <bebarino@gmail.com>
> [...]
>
> Encapsulation is good.
>
> [..]
>
> This is IMHO slightly less readable, and doesn't protect against
> changes in implementation.

Agreed.

>>       for (i = 0; i < rename_src_nr; i++)
>>               insert_file_table(&file_table, -1, i, rename_src[i].one);
>>
>> diff --git a/hash.h b/hash.h
>> index 69e33a4..418be24 100644
>> --- a/hash.h
>> +++ b/hash.h
>> @@ -33,11 +33,4 @@ extern void **insert_hash(unsigned int hash, void *ptr, struct hash_table *table
>>  extern int for_each_hash(const struct hash_table *table, int (*fn)(void *));
>>  extern void free_hash(struct hash_table *table);
>>
>> -static inline void init_hash(struct hash_table *table)
>> -{
>> -     table->size = 0;
>> -     table->nr = 0;
>> -     table->array = NULL;
>> -}
>
> *This* could be replaced by memset.

No it couldn't? The second argument to memset is just an int, so
setting the memory area to 0 isn't portable to systems where the
representation of NULL isn't "0".

(It's early so I may be misremembering my C..)

Re: [PATCH] hash: Remove useless init_hash()

[Jeff King]

On Tue, Jul 27, 2010 at 10:36:09AM +0000, Ævar Arnfjörð Bjarmason wrote:

> >> -static inline void init_hash(struct hash_table *table)
> >> -{
> >> -     table->size = 0;
> >> -     table->nr = 0;
> >> -     table->array = NULL;
> >> -}
> >
> > *This* could be replaced by memset.
> 
> No it couldn't? The second argument to memset is just an int, so
> setting the memory area to 0 isn't portable to systems where the
> representation of NULL isn't "0".
> 
> (It's early so I may be misremembering my C..)

You're remembering your C correctly. It isn't portable, but it is so
unlikely on modern machines that we simply don't care (and you will see
memsets zero-ing pointers like this all through the git code, so this is
certainly not introducing anything new).

That being said, I agree with the comments that removing init_hash
actually makes the code _less_ readable. You could just replace these
three lines with a memset, but why? It's just code churn.

-Peff

Re: [PATCH] hash: Remove useless init_hash()

[Ævar Arnfjörð Bjarmason]

On Tue, Jul 27, 2010 at 19:49, Jeff King <peff@peff.net> wrote:
> On Tue, Jul 27, 2010 at 10:36:09AM +0000, Ævar Arnfjörð Bjarmason wrote:
>
>> >> -static inline void init_hash(struct hash_table *table)
>> >> -{
>> >> -     table->size = 0;
>> >> -     table->nr = 0;
>> >> -     table->array = NULL;
>> >> -}
>> >
>> > *This* could be replaced by memset.
>>
>> No it couldn't? The second argument to memset is just an int, so
>> setting the memory area to 0 isn't portable to systems where the
>> representation of NULL isn't "0".
>>
>> (It's early so I may be misremembering my C..)
>
> You're remembering your C correctly. It isn't portable, but it is so
> unlikely on modern machines that we simply don't care (and you will see
> memsets zero-ing pointers like this all through the git code, so this is
> certainly not introducing anything new).

Thanks for the confirmation. I was aware that NULL != 0 only occured
on long-dead architechtures, but I suspected that some compiler out
there would whine if it could statically determine that you were
reading in a memzero'd area and using it as NULL. Evidently not, or at
least nobody's complained.

> That being said, I agree with the comments that removing init_hash
> actually makes the code _less_ readable. You could just replace these
> three lines with a memset, but why? It's just code churn.

Yeah, and for the record it also missed this part in hash.c:

    void free_hash(struct hash_table *table)
    {
    	free(table->array);
    	table->array = NULL;
    	table->size = 0;
    	table->nr = 0;
    }

Have fun everyone :)

Re: [PATCH] hash: Remove useless init_hash()

[Stephen Boyd]

  On 07/27/2010 12:58 PM, Ævar Arnfjörð Bjarmason wrote:
>
>> That being said, I agree with the comments that removing init_hash
>> actually makes the code _less_ readable. You could just replace these
>> three lines with a memset, but why? It's just code churn.
> Yeah, and for the record it also missed this part in hash.c:
>
>      void free_hash(struct hash_table *table)
>      {
>      	free(table->array);
>      	table->array = NULL;
>      	table->size = 0;
>      	table->nr = 0;
>      }
>
> Have fun everyone :)

Ok, seems like nobody thinks this is a good idea so I'm fine with dropping it. The minimal savings (if at all) doesn't seem worth the code churn. One less patch in the queue ;-)

Re: [PATCH] hash: Remove useless init_hash()

[Linus Torvalds]

On Mon, Jul 26, 2010 at 10:36 PM, Stephen Boyd <bebarino@gmail.com> wrote:
>
> init_hash() is essentially a memset() so just use that.

I disagree.

Yes, technically right now the initialization just zeroes out all the
fields, and memset() does the same. But there is no advantage to using
memset(), since init_hash() will create the same or better code, and
using init_hash() is way more readable.

Also, it's not at all the case that init_hash() is always going to be
a memset(). Imagine a threaded hash-table with a lock associated with
it or something. Now, admittedly that's not necessarily something we'd
ever do in git, but I still think it's simply a good idea to have a
clear "initialize this" routine, rather than depending on the fact
that zeroing it out is sufficient.

                    Linus