Re: st: macro problem

[Nick Cox <njcoxstata@gmail.com>]

Note first that I didn't recommend either of your examples; and second
-- and much more importantly -- that they make more sense when you
follow the principles I outlined.

When you define

local list "a b" "b c" "c d"

Stata strips the outermost " " as delimiters so (for example) "a b" is
no longer a single word (Stata sense), but two words a and b". Stata
evidently has rules that differ for single " on the left and the
right, but the overarching principle remains that the way to insist
that " " remain as such is to wrap them in further delimiters.

Nick

On Mon, Jun 11, 2012 at 8:04 PM, Jeph Herrin <stata@spandrel.net> wrote:
> Thanks. In both cases I should have clarified that I want the list of match
> strings to be part of a macro (as in my example) because the list will
> change from time to time (and I prefer to isolate code changes to macros).
> In that case, your example and others produce surprising results:
>
>  . local list "a b" "b c" "c d"
>
>  . foreach s in `list' {
>
>   2.    di "`s'"
>   3.    }
>  a
>  b
>  b c
>  c d
>
> . local list "a" "b"
>
> . foreach s in `list' {
>
>  2.    di `"`s'"'
>  3.    }
>  a"
>  b
>
> (note the trailing " on the -a-) and so on.
>
> cheers,
> Jeph
>
>
>
>
>
>
> On 6/11/2012 12:29 PM, Nick Cox wrote:
>>
>> " " have two roles in Stata; they act as string delimiters (which by
>> default are stripped) and they act as literal characters which you want to
>> preserve as part of a string (in which you often need `" "' as delimiters).
>>
>> However, this works
>>
>> . foreach s in "a b" "b c" "c d" {
>>   2. di "`s'"
>>   3. }
>> a b
>> b c
>> c d
>>
>> so maybe something simpler would also work for you.
>>
>> Nick
>> n.j.cox@durham.ac.uk
>>
>> Jeph Herrin
>>
>> There is no problem with -strpos-, which never returns missing and which
>> is zero exactly when I want it to be. The problem was in creating the
>> original macro and then referencing the elements so that they appear in
>> -strpos- with the correct number of quotations marks; I was achieving
>> either too many or too few of the latter.
>>
>> The solution (or at least, a solution) is for local -matchlist- needed
>> to be enclosed in compound quotes, and then ditto the reference:
>>
>>     local matchlist `""string 1" "string 2" "string 3" .... "string 55""'
>>
>>     gen byte match=0
>>     foreach S of local matchlist {
>>          replace match = 1 if strpos(strvar,`"`S'"')
>>     }
>>
>>
>> The reference in -strpos- looks wrong to me - I would anticipate that
>> `S' would include the necessary "'s for -strpos-, but it does not.
>> Without the extra quotes on the original local, however, `"`S'"'
>> resolves to include too many "'s.
>>
>> On 6/11/2012 11:58 AM, Richard Goldstein wrote:
>>
>>> you don't make it clear about what doesn't work
>>>
>>> but here is a guess: you are getting match=1 in all cases (or at least
>>> all where strvar is not missing); remember, however, that strpos=0 if
>>> "S" is not found; so, in addition to the quotes, etc., I guess you want
>>> to ensure that strpos is>0 (and maybe<.)
>>
>>
>> On 6/11/12 11:50 AM, Jeph Herrin wrote:
>>
>>>> This should be trivial, but for some reason I cannot get it.
>>>>
>>>> I have a string variable, and I have a list of string values I need to
>>>> match against it. There are occasional trailing or leading characters in
>>>> the variable, so I am using -strpos- to find matches. Thus
>>>>
>>>>    local matchlist "string 1" "string 2"  "string 3" .... "string 55"
>>>>
>>>>    gen byte match=0
>>>>    foreach S in `matchlist' {
>>>>          replace match = 1 if strpos(strvar,`S')
>>>>    }
>>>>
>>>> Now, I know I have got the quotes and macro evaluations wrong, but that
>>>> is the problem: I've tried this many different ways, and yet cannot seem
>>>> to find a combination that works. Any thoughts?
>>

*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/