[Rd] Improving string concatenation
maechler at stat.math.ethz.ch
Fri Jun 19 12:31:28 CEST 2015
>>>>> Radford Neal <radford at cs.toronto.edu>
>>>>> on Thu, 18 Jun 2015 14:32:18 -0400 writes:
> Gabor Csardi writes:
>> Btw. for some motivation, here is a (surely incomplete)
>> list of languages with '+' as the string concatenation
>> ALGOL 68, BASIC, C++, C#, Cobra, Pascal, Object Pascal,
>> Windows Powers hell, Objective-C, F#, Sc-ala, Ya.
> The situation for R is rather different from that of a
> language (as many of the above) in which variables are
> declared to be of a specific type.
> In such a statically typed language, when you see the
> expression "a+b", it is easy to figure out whether the "+"
> will be numeric addition or string concatenation, by
> looking at the declarations of "a" and "b".
> But in a language such as R in which values have types,
> but variables don't, someone seeing "a+b" in code wouldn't
> be able to tell easily what it does. This is OK, in fact
> desirable, in the case of operator dispatch according to
> class when the different methods implement versions of the
> operator that have analogous properties. But numeric
> addition and string concatenation have just about nothing
> in common, so cases where functions are meant to have "+"
> be either addition OR concatenation are going to be rare.
> Furthermore, making "+" concatenate strings would preclude
> ever making "+" convert strings to numbers (signalling an
> error if they aren't in some numerical format) and then
> add them. I'm not sure whether that would be a good idea
> or not, but it might be unwise to eliminate the
> And of course, as someone else mentioned, it may instead
> be desirable for attempts to add strings to signal an
> error, as at present, which one also gives up by making
> "+" do concatenation.
>> Yes, even Fortran has one, and in C, I can simply write
>> "literal1" "literal2" and they'll be concatenated. It is
>> only for literals, but still very useful.
> Concatenation of literal strings could easily be added to
> the R parser without changing anything else. (Getting
> them to deparse as the same two pieces would be tricky,
> but is maybe not necessary.)
> Radford Neal
Thank you, Bill Dunlap, Radford, Herv'e,
and others who have explained indirectly that the subject of this thread
is overall rather incomplete or just not true:
Such "improvemnet" -- making something more convenient in many
cases -- would lead to (backward) incompatibilities -- breaking
current functionality -- and inconsistencies in R.
As this thread hopefully comes to a conclusion for now,
let me try mention that this is not the first time the topic has
come up ... and those of us who may still stick around in 5
years, please try to remember : It will come up every few
Nine years ago was one such occasion --- on this same place,
Here, I've started it (as Friday afternoon "event", diverting
from the more relevant topic of S4 methods for "+"),
see also here
or the Gmane archive version of it, e.g. with John Chambers (citing Bill Dunlap)
Also, the arguments (against "+" for string concatenation)
of Thomas Lumley were *not* repeated this time (I think)
Historical note: At the time, there was only 'paste()' in base R.
Gabor did mention paste0() as a possible compromise,
and indeed, we did add paste0() to R eventually.
Maybe we should make this into a new R-FAQ entry -- so next
time, we can point there instead of re-hashing things ever so often.
I'm volunteering to collect "patches" -- ideally texinfo format,
the latest source of the R FAQ list being
More information about the R-devel