Xapian devel - Dec 2014 - Replace atoi and atol with strtol strtoul:Need Help

Hello ,
I came across this function *HtmlParser::decode_entities(string &s)* in
*xapian-application/omega/htmlparse.cc* which basically does is extract hex
value if any or extract number.For extracting number atoi is used and value
returned by it is stored in variable "val" , I think so replacing atoi
with
strtoul would be useful here as number can have larger value although the
variable "val" is unsigned int so i need to change the that definition
of
"val" also. Is that ok to do so ? Just need to clarify .

On 16 December 2014 at 04:34, Olly Betts <olly at survex.com>
wrote:
>
> On Tue, Dec 16, 2014 at 02:32:31AM +0530, Priyank Bhatt wrote:
> > I am working on replacing atoi () and atol() functions with strtol()
and
> > strtoul() . I came across many files which uses statement  like these
> > time_t secs= atoi(data_span.c_str()), here time_t Datatype is not
known
> but
> > wikipedia says that it is integer so is it necessary to replace  atoi
> with
> > strtol over here ??
>
> The time_t type is a standard one - ISO C only says it's a
"arithmetic
> type" (so it could potentially be a double) but POSIX says it's an
> "integer type":
>
> http://pubs.opengroup.org/onlinepubs/9699919799/basedefs/sys_types.h.html
>
> While that would allow it to be unsigned, in practice it seems to be a
> signed integer, and it is 64 bit on modern systems (since a signed 32
> bit integer with the Unix epoch can only represent dates up to 2038).
>
> You missed out a relevant part of line, which in full is:
>
> xapian-applications/omega/date.cc:      time_t secs >
atoi(date_span.c_str()) * (24 * 60 * 60);
>
> So in this case, date_span is a number of days, so converting it via a
> long is reasonable - even if long is 32 bits that can still represent a
> span of 5.8 million years.  You'd want to make sure the multiplication
> happens in type time_t though.
>
> If we were actually converting a number of seconds, you'd want to use
> strtoll(), at least if sizeof(long) < 8.  C++11 includes strtoll(), and
> we decided to require a C++11 compiler for 1.3.x, so there's no need to
> worry whether strtoll() is available.
>
> > And is their any document which helps me what each file function does
> like
> > date.cc,cgiparams.cc,omega.cc,query.cc.
>
> There should be a short comment summarising the purpose of each file at
> the top, for example:
>
> | /* date.cc: date range parsing routines for omega
>
> A few files may still lack these, but most have them.
>
> Cheers,
>     Olly
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
<http://lists.xapian.org/pipermail/xapian-devel/attachments/20141217/ec869922/attachment-0002.html>

On Wed, Dec 17, 2014 at 01:15:12AM +0530, Priyank Bhatt
wrote:
> I came across this function *HtmlParser::decode_entities(string &s)* in
> *xapian-application/omega/htmlparse.cc* which basically does is extract hex
> value if any or extract number.
The code you refer to is actually parsing a decimal value (like &) -
the hex case (like &) uses sscanf().
> For extracting number atoi is used and value
> returned by it is stored in variable "val" , I think so replacing
atoi with
> strtoul would be useful here as number can have larger value although the
> variable "val" is unsigned int so i need to change the that
definition of
> "val" also. Is that ok to do so ? Just need to clarify .
We ultimately pass val to Xapian::Unicode::nonascii_to_utf8() which
takes "unsigned" so there's not much point making val a wider type
here.

While ISO C/C++ only guarantee that int is at least 16 bits, in
practice it is 32 bits on the platforms we support.

Cheers,
    Olly

Hello,
I came across the file *omega.cc* which is in directory*
xapain-application/omega/*

In this file , atoi is used in *Percentage Relevance cutoff *(293 line no)
as Percentage lies between 0-100 their is no need to modify atoi . But do
we need to check for error's ?

Second Implementation is in *collapsing* (301)  in which we collapse set of
document under a key,range of this key has not been defined anywhere so we
can increase the size of the key over here to accumulate more key's ?

Third Implementation is in *Sort* (330) And I am not sure what is the val
value over here is ??  Is it the entire string of sorted value number's as
cgi_params is multimap and find returns the iterator at which it find the
element containing the key value ? And I am not sure whether to modify atoi
over here.

val = cgi_params.find("SORT");
    if (val != cgi_params.end()) {
    sort_key = atoi(val->second.c_str());

Thank You,
Priyank Bhatt

On 17 December 2014 at 03:38, Olly Betts <olly at survex.com>
wrote:
>
> On Wed, Dec 17, 2014 at 01:15:12AM +0530, Priyank Bhatt wrote:
> > I came across this function *HtmlParser::decode_entities(string
&s)* in
> > *xapian-application/omega/htmlparse.cc* which basically does is
extract
> hex
> > value if any or extract number.
>
> The code you refer to is actually parsing a decimal value (like &#38;)
-
> the hex case (like &#x26;) uses sscanf().
>
> > For extracting number atoi is used and value
> > returned by it is stored in variable "val" , I think so
replacing atoi
> with
> > strtoul would be useful here as number can have larger value although
the
> > variable "val" is unsigned int so i need to change the that
definition of
> > "val" also. Is that ok to do so ? Just need to clarify .
>
> We ultimately pass val to Xapian::Unicode::nonascii_to_utf8() which
> takes "unsigned" so there's not much point making val a wider
type
> here.
>
> While ISO C/C++ only guarantee that int is at least 16 bits, in
> practice it is 32 bits on the platforms we support.
>
> Cheers,
>     Olly
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
<http://lists.xapian.org/pipermail/xapian-devel/attachments/20141219/2f50e170/attachment-0002.html>