[Bioc-devel] depends on packages providing classes
What is the current best paradigm for using all the classes in S4Vectors/GenomeInfoDb/GenomicRanges/IRanges I obviously import methods and classes from the relevant packages. But shouldn't I depend on these packages as well? Since I basically want the user to have this functionality at the command line? That is what I do now. That of course leads to the R CMD check NOTE on depending on too many packages I guess I should ignore that one. Best, Kasper [[alternative HTML version deleted]] ___ Bioc-devel@r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/bioc-devel
Re: [Bioc-devel] depends on packages providing classes
Hi, On 10/28/2014 08:48 AM, Vincent Carey wrote: On Tue, Oct 28, 2014 at 11:23 AM, Kasper Daniel Hansen kasperdanielhan...@gmail.com wrote: Well, first I want to make sure that there is not something special regarding S4 methods and classes. I have a feeling that they are a special case. Second, while I agree with Jim's general opinion, it is a little bit different when I have return objects which are defined in other packages. If I don't depend on this other package, the user is hosed wrt. the return object, unless I manually export all classes from this other In what sense? If you return an instance of GRanges, certain things can be done even if GenomicRanges is not attached. Yes certain things maybe, but it's hard to predict which ones. You can get values of slots, for example. With the following little package %vjcair cat foo/NAMESPACE importFrom(IRanges, IRanges) importClassesFrom(GenomicRanges, GRanges) importFrom(GenomicRanges, GRanges) export(myfun) %vjcair cat foo/DESCRIPTION Package: foo Title: foo Version: 0.0.0 Author: VJ Carey st...@channing.harvard.edu Description: Suggests: Depends: Imports: GenomicRanges Maintainer: VJ Carey st...@channing.harvard.edu License: Private LazyLoad: yes %vjcair cat foo/R/* myfun = function(seqnames=1, ranges=IRanges(1,2), ...) GRanges(seqnames=seqnames, ranges=ranges, ...) The following works: library(foo) x = myfun() x GRanges object with 1 range and 0 metadata columns: seqnamesranges strand Rle IRanges Rle [1]1[1, 2] * --- seqinfo: 1 sequence from an unspecified genome; no seqlengths So the show method works, even though I have not touched it. (I did not expect it to work, in fact.) Exactly. Let's call it luck ;-) Additionally, I can get access to slots. The end user should never try to access slots directly but use getters and setters instead. And most getters and setters for GRanges objects are defined and documented in the GenomicRanges package. Those that are not are defined in packages that GenomicRanges depends on. But ranges() fails. If I, the user, want to use it, I need to arrange for that. IMO if your package returns a GRanges object to the user, then the user should be able to access the man page for GRanges objects with ?GRanges. And that works only if the GenomicRanges package is attached. Attaching GenomicRanges will also attach other packages that GenomicRanges depends on where some GRanges accessors might be defined and documented (e.g. metadata()). In some cases you'll decide you want the user to have a full complement of methods for your package to function meaningfully. For example, I am considering using dplyr idioms to work with data structures in a package, and it seems I should just depend on dplyr rather than pick out and document which things I want to expose. But that may still be an undesirable design. package, like importClassesFrom(GenomicRanges, GRanges) exportClasses(GRanges) Surely that is not intended. It is important that my package works without being attached to the search path and I do this by carefully importing what I need, ie. my code does not require that my dependencies are attached to the search path. But the end user will be hosed without it. Yes s/he will. Fortunately when your package namespace gets loaded by another package, then nothing gets attached to the search path, even if your package depends (instead of imports) on other packages. So using Depends instead of Imports for your own dependencies won't make any difference in that respect, which is good. My impression is that the NOTE in R CMD check was written by someone who did not anticipate large-scale use and re-use of classes and methods across many packages. That's my impression too. Cheers, H. Best, Kasper On Tue, Oct 28, 2014 at 11:14 AM, James W. MacDonald jmac...@uw.edu wrote: I agree with Vince. It's your job as a package developer to make available to your package all the functions necessary for the package to work. But I am not sure it is your job to load all the packages that your end user might need. Best, Jim On Tue, Oct 28, 2014 at 11:04 AM, Vincent Carey st...@channing.harvard.edu wrote: On Tue, Oct 28, 2014 at 10:19 AM, Kasper Daniel Hansen kasperdanielhan...@gmail.com wrote: What is the current best paradigm for using all the classes in S4Vectors/GenomeInfoDb/GenomicRanges/IRanges I obviously import methods and classes from the relevant packages. But shouldn't I depend on these packages as well? Since I basically want the user to have this functionality at the command line? That is what I do now. I've wondered about this as well. It seems the principle is that the user should take care of attaching additional packages when needed. It might be appropriate to give a hint in the package startup message, if having some other package attached would
Re: [Bioc-devel] depends on packages providing classes
On Tue, Oct 28, 2014 at 2:29 PM, Hervé Pagès hpa...@fredhutch.org wrote: Hi, On 10/28/2014 08:48 AM, Vincent Carey wrote: On Tue, Oct 28, 2014 at 11:23 AM, Kasper Daniel Hansen kasperdanielhan...@gmail.com wrote: Well, first I want to make sure that there is not something special regarding S4 methods and classes. I have a feeling that they are a special case. Second, while I agree with Jim's general opinion, it is a little bit different when I have return objects which are defined in other packages. If I don't depend on this other package, the user is hosed wrt. the return object, unless I manually export all classes from this other In what sense? If you return an instance of GRanges, certain things can be done even if GenomicRanges is not attached. Yes certain things maybe, but it's hard to predict which ones. You can get values of slots, for example. With the following little package %vjcair cat foo/NAMESPACE importFrom(IRanges, IRanges) importClassesFrom(GenomicRanges, GRanges) importFrom(GenomicRanges, GRanges) export(myfun) %vjcair cat foo/DESCRIPTION Package: foo Title: foo Version: 0.0.0 Author: VJ Carey st...@channing.harvard.edu Description: Suggests: Depends: Imports: GenomicRanges Maintainer: VJ Carey st...@channing.harvard.edu License: Private LazyLoad: yes %vjcair cat foo/R/* myfun = function(seqnames=1, ranges=IRanges(1,2), ...) GRanges(seqnames=seqnames, ranges=ranges, ...) The following works: library(foo) x = myfun() x GRanges object with 1 range and 0 metadata columns: seqnamesranges strand Rle IRanges Rle [1]1[1, 2] * --- seqinfo: 1 sequence from an unspecified genome; no seqlengths So the show method works, even though I have not touched it. (I did not expect it to work, in fact.) Exactly. Let's call it luck ;-) Additionally, I can get access to slots. The end user should never try to access slots directly but use getters and setters instead. And most getters and setters for GRanges objects are defined and documented in the GenomicRanges package. Those that are not are defined in packages that GenomicRanges depends on. But ranges() fails. If I, the user, want to use it, I need to arrange for that. IMO if your package returns a GRanges object to the user, then the user should be able to access the man page for GRanges objects with ?GRanges. Oddly enough, that seems to be incorrect. I added a man page to foo that has a \link[GenomicRanges]{GRanges-class}. I ran help.start and the cross reference from my man page succeeds. Furthermore with the sessionInfo below, ?GRanges succeeds at the CLI. I am not trying to defend the NOTE but the principle of minimizing Depends declarations needs to be considered critically, and I am just exploring the space. ?GRanges # it worked as usual in the tty sessionInfo() R version 3.1.1 (2014-07-10) Platform: x86_64-apple-darwin13.1.0 (64-bit) locale: [1] en_US.UTF-8/en_US.UTF-8/en_US.UTF-8/C/en_US.UTF-8/en_US.UTF-8 attached base packages: [1] stats graphics grDevices datasets utils tools methods [8] base other attached packages: [1] foo_0.0.0rmarkdown_0.3.8 knitr_1.6 [4] weaver_1.31.0codetools_0.2-9 digest_0.6.4 [7] BiocInstaller_1.16.0 loaded via a namespace (and not attached): [1] BiocGenerics_0.11.5 evaluate_0.5.5formatR_1.0 [4] GenomeInfoDb_1.1.26 GenomicRanges_1.17.48 htmltools_0.2.6 [7] IRanges_1.99.32 parallel_3.1.1S4Vectors_0.2.8 [10] stats4_3.1.1 stringr_0.6.2 XVector_0.5.8 And that works only if the GenomicRanges package is attached. Attaching GenomicRanges will also attach other packages that GenomicRanges depends on where some GRanges accessors might be defined and documented (e.g. metadata()). In some cases you'll decide you want the user to have a full complement of methods for your package to function meaningfully. For example, I am considering using dplyr idioms to work with data structures in a package, and it seems I should just depend on dplyr rather than pick out and document which things I want to expose. But that may still be an undesirable design. package, like importClassesFrom(GenomicRanges, GRanges) exportClasses(GRanges) Surely that is not intended. It is important that my package works without being attached to the search path and I do this by carefully importing what I need, ie. my code does not require that my dependencies are attached to the search path. But the end user will be hosed without it. Yes s/he will. Fortunately when your package namespace gets loaded by another package, then nothing gets attached to the search path, even if your package depends (instead of imports) on other packages. So using Depends instead of Imports for your own dependencies won't make
Re: [Bioc-devel] depends on packages providing classes
On 10/28/2014 12:42 PM, Vincent Carey wrote: On Tue, Oct 28, 2014 at 2:29 PM, Hervé Pagès hpa...@fredhutch.org mailto:hpa...@fredhutch.org wrote: Hi, On 10/28/2014 08:48 AM, Vincent Carey wrote: On Tue, Oct 28, 2014 at 11:23 AM, Kasper Daniel Hansen kasperdanielhan...@gmail.com mailto:kasperdanielhan...@gmail.com wrote: Well, first I want to make sure that there is not something special regarding S4 methods and classes. I have a feeling that they are a special case. Second, while I agree with Jim's general opinion, it is a little bit different when I have return objects which are defined in other packages. If I don't depend on this other package, the user is hosed wrt. the return object, unless I manually export all classes from this other In what sense? If you return an instance of GRanges, certain things can be done even if GenomicRanges is not attached. Yes certain things maybe, but it's hard to predict which ones. You can get values of slots, for example. With the following little package %vjcair cat foo/NAMESPACE importFrom(IRanges, IRanges) importClassesFrom(__GenomicRanges, GRanges) importFrom(GenomicRanges, GRanges) export(myfun) %vjcair cat foo/DESCRIPTION Package: foo Title: foo Version: 0.0.0 Author: VJ Carey st...@channing.harvard.edu mailto:st...@channing.harvard.edu Description: Suggests: Depends: Imports: GenomicRanges Maintainer: VJ Carey st...@channing.harvard.edu mailto:st...@channing.harvard.edu License: Private LazyLoad: yes %vjcair cat foo/R/* myfun = function(seqnames=1, ranges=IRanges(1,2), ...) GRanges(seqnames=seqnames, ranges=ranges, ...) The following works: library(foo) x = myfun() x GRanges object with 1 range and 0 metadata columns: seqnamesranges strand Rle IRanges Rle [1]1[1, 2] * --- seqinfo: 1 sequence from an unspecified genome; no seqlengths So the show method works, even though I have not touched it. (I did not expect it to work, in fact.) Exactly. Let's call it luck ;-) Additionally, I can get access to slots. The end user should never try to access slots directly but use getters and setters instead. And most getters and setters for GRanges objects are defined and documented in the GenomicRanges package. Those that are not are defined in packages that GenomicRanges depends on. But ranges() fails. If I, the user, want to use it, I need to arrange for that. IMO if your package returns a GRanges object to the user, then the user should be able to access the man page for GRanges objects with ?GRanges. Oddly enough, that seems to be incorrect. I added a man page to foo that has a \link[GenomicRanges]{GRanges-class}. I ran help.start and the cross reference from my man page succeeds. Furthermore with the sessionInfo below, ?GRanges succeeds at the CLI. Did you try to run example(GRanges)? I'm not sure that will work. For example after I do library(rtracklayer), I can indeed do ?DNAStringSet at the command line (I'm surprised this works), but then example(DNAStringSet) fails: example(DNAStringSet) Warning message: In example(DNAStringSet) : no help found for ‘DNAStringSet’ I'm also surprised this is just a warning but that's another story... H. I am not trying to defend the NOTE but the principle of minimizing Depends declarations needs to be considered critically, and I am just exploring the space. ?GRanges # it worked as usual in the tty sessionInfo() R version 3.1.1 (2014-07-10) Platform: x86_64-apple-darwin13.1.0 (64-bit) locale: [1] en_US.UTF-8/en_US.UTF-8/en_US.UTF-8/C/en_US.UTF-8/en_US.UTF-8 attached base packages: [1] stats graphics grDevices datasets utils tools methods [8] base other attached packages: [1] foo_0.0.0rmarkdown_0.3.8 knitr_1.6 [4] weaver_1.31.0codetools_0.2-9 digest_0.6.4 [7] BiocInstaller_1.16.0 loaded via a namespace (and not attached): [1] BiocGenerics_0.11.5 evaluate_0.5.5formatR_1.0 [4] GenomeInfoDb_1.1.26 GenomicRanges_1.17.48 htmltools_0.2.6 [7] IRanges_1.99.32 parallel_3.1.1S4Vectors_0.2.8 [10] stats4_3.1.1 stringr_0.6.2 XVector_0.5.8 And that works only if the GenomicRanges package is attached. Attaching GenomicRanges will also attach other packages that GenomicRanges depends
[Rd] RFC: is.whole() ? {how to judge if a variable is an integer}
Diverted to R-devel, as I'm requesting comments about a proposal to add is.whole() to R just so this issue does not trail on for centuries (;-), see below. William Dunlap wdun...@tibco.com on Sat, 18 Oct 2014 10:33:05 -0700 writes: 3. all.equal(a, as.integer(a)) Note that this one tests if 'a' can be stored accurately as a 32-bit signed integer. If you want to know if 'a' can be used as an accurate count, then you want to test if a+1a (use abs() in case a is negative). E.g., try this for a-2^49-1, about 5*10^14. You have to decide what properties of integers you are interested in. Bill Dunlap TIBCO Software wdunlap tibco.com good point, thank you, Bill. This whole issue comes up about once year (at least), it seems to me, and every time there are some good and some not so good propositions, some aiming for vectorized / whole object solutions some, only assuming scalar input. Indeed, I'd claim that all three proposals cited below are not good enough in one way or the other though one could argue that a tolerance = 0 version would be good enough, and hence the function(x) x %% 1 == 0 would be sufficient. In the CRAN package 'sfsmisc' (which I maintain), Alain Hauser recently added an is.whole() function which works vectorized and uses a 'tolerance' and then all.equal(), but in better way than (most / all ?) what was proposed in this thread (line 486 ff of https://github.com/mmaechler/sfsmisc/blob/master/R/misc-goodies.R ). Further, the CRAN packages 'gmp' (infinite precision integer and rational arithmetic) and 'Rmpfr' ('infinite' precision floating point arithmetic) contain an S3 generic is.whole() function and methods for their own number classes and a default method (( which however assumes a tolerance of zero, the same as the x %% 1 == 0 proposal)). Given all this, the r-help and stackoverflow threads, maybe we should decide that such an is.whole() function should be added to R and maintained (by me for the time), so we do have a better place to point people to, and well documented -- and eventually optimized -- behavior ? Personally I do think I'd want a signature of function (x, tolerance = sqrt(.Machine$double.eps)) Martin On Sat, Oct 18, 2014 at 10:02 AM, PO SU rhelpmaill...@163.com wrote: Tks for your help, after investigate in your link, i find there seems three ways can be adoped: 1.is.wholenumber - function(x, tol = .Machine$double.eps^0.5) abs(x - round(x)) tol) e.g. is.wholenumber(1) 2. x%%1==0 3. all.equal(a, as.integer(a)) and also included your last suggestion using floor. and also tks for other helpers! -- PO SU mail: desolato...@163.com Majored in Statistics from SJTU At 2014-10-18 22:48:15, Sergio Fonda sergio.fond...@gmail.com wrote: Sorry for my previous hurry misunderstanding. Try this link: http://stackoverflow.com/questions/3476782/how-to-check-if-the-number-is-integer 2014-10-18 16:25 GMT+02:00 PO SU rhelpmaill...@163.com: It's due to that, 1 is a numeric, 1.2 is a numeric, though it's true. but deeply, when i want to know 1 is an integer, there seems no easy way to get the answer. So, is there anyone happen to know it? At 2014-10-18 20:10:09, S Ellison s.elli...@lgcgroup.com wrote: But i use a-10/b , b is some value ,may be 5, maybe 5.5 If you do floating point arithmetic on integers you'll usually get floating point answers, including the 5.0. See FAQ 7.31 for the usual floating point problem, and ?all.equal for the usual answer to it. You could see if a result is close to an integer by,for example, using all.equal to compare it to itself after rounding. S __ R-devel@r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-devel
Re: [Rd] RFC: is.whole() ? {how to judge if a variable is an integer}
On 28/10/2014, 5:32 AM, Martin Maechler wrote: Diverted to R-devel, as I'm requesting comments about a proposal to add is.whole() to R just so this issue does not trail on for centuries (;-), see below. I didn't read the thread in question, but I think Bill's comment is crucial: what properties of integers are you interested in testing? One that comes up a lot is that some non-integral values print as integers, and people get confused when using them as indices. For example, this prints results that surprise some people: x - 1:10 index - 4 - 1.e-10 index x[index] I would think is.whole(index) should return FALSE here, but the sfsmisc function defaults to a nonzero tolerance, and returns TRUE. Duncan Murdoch William Dunlap wdun...@tibco.com on Sat, 18 Oct 2014 10:33:05 -0700 writes: 3. all.equal(a, as.integer(a)) Note that this one tests if 'a' can be stored accurately as a 32-bit signed integer. If you want to know if 'a' can be used as an accurate count, then you want to test if a+1a (use abs() in case a is negative). E.g., try this for a-2^49-1, about 5*10^14. You have to decide what properties of integers you are interested in. Bill Dunlap TIBCO Software wdunlap tibco.com good point, thank you, Bill. This whole issue comes up about once year (at least), it seems to me, and every time there are some good and some not so good propositions, some aiming for vectorized / whole object solutions some, only assuming scalar input. Indeed, I'd claim that all three proposals cited below are not good enough in one way or the other though one could argue that a tolerance = 0 version would be good enough, and hence the function(x) x %% 1 == 0 would be sufficient. In the CRAN package 'sfsmisc' (which I maintain), Alain Hauser recently added an is.whole() function which works vectorized and uses a 'tolerance' and then all.equal(), but in better way than (most / all ?) what was proposed in this thread (line 486 ff of https://github.com/mmaechler/sfsmisc/blob/master/R/misc-goodies.R ). Further, the CRAN packages 'gmp' (infinite precision integer and rational arithmetic) and 'Rmpfr' ('infinite' precision floating point arithmetic) contain an S3 generic is.whole() function and methods for their own number classes and a default method (( which however assumes a tolerance of zero, the same as the x %% 1 == 0 proposal)). Given all this, the r-help and stackoverflow threads, maybe we should decide that such an is.whole() function should be added to R and maintained (by me for the time), so we do have a better place to point people to, and well documented -- and eventually optimized -- behavior ? Personally I do think I'd want a signature of function (x, tolerance = sqrt(.Machine$double.eps)) Martin On Sat, Oct 18, 2014 at 10:02 AM, PO SU rhelpmaill...@163.com wrote: Tks for your help, after investigate in your link, i find there seems three ways can be adoped: 1.is.wholenumber - function(x, tol = .Machine$double.eps^0.5) abs(x - round(x)) tol) e.g. is.wholenumber(1) 2. x%%1==0 3. all.equal(a, as.integer(a)) and also included your last suggestion using floor. and also tks for other helpers! -- PO SU mail: desolato...@163.com Majored in Statistics from SJTU At 2014-10-18 22:48:15, Sergio Fonda sergio.fond...@gmail.com wrote: Sorry for my previous hurry misunderstanding. Try this link: http://stackoverflow.com/questions/3476782/how-to-check-if-the-number-is-integer 2014-10-18 16:25 GMT+02:00 PO SU rhelpmaill...@163.com: It's due to that, 1 is a numeric, 1.2 is a numeric, though it's true. but deeply, when i want to know 1 is an integer, there seems no easy way to get the answer. So, is there anyone happen to know it? At 2014-10-18 20:10:09, S Ellison s.elli...@lgcgroup.com wrote: But i use a-10/b , b is some value ,may be 5, maybe 5.5 If you do floating point arithmetic on integers you'll usually get floating point answers, including the 5.0. See FAQ 7.31 for the usual floating point problem, and ?all.equal for the usual answer to it. You could see if a result is close to an integer by,for example, using all.equal to compare it to itself after rounding. S __ R-devel@r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-devel __ R-devel@r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-devel
Re: [Rd] is.whole()
Martin, I can't imagine using such a function myself, the reason being that as Bill and Duncan point out, the correct answer depends on the situation. But given the regular reappearance of this topic, I think that perhaps creation of your function is a good idea, largely to function as a repository for the knowlege. If one takes that view, then perhas the function has two optional arguments: case and tolerance. The first would choose a scenario of exact, numeric, count, etc, where exact refers to Duncan's case, numeric to your default, and count to Bill's a+1 a. The second argument would be rarely used. The primary point of the function would be the Details section of its manual page. Whenver the issue comes up the response could then be see the is.whole() function and its documentation. Terry T. On 10/28/2014 06:00 AM, r-devel-requ...@r-project.org wrote: Diverted to R-devel, as I'm requesting comments about a proposal to add is.whole() to R just so this issue does not trail on for centuries (;-), see below. __ R-devel@r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-devel
Re: [Rd] is.whole()
Just to anticipate future discussion from math purists (and hopefully not to throw too much of a wrench in the works), what would be the return of: is.whole(-1) or is.whole(-1L) ? I can see arguments for both TRUE and FALSE from both the math purity group and the what will happen when I try to use this for subsetting? group. On Tue, Oct 28, 2014 at 9:14 AM, Martin Maechler maech...@stat.math.ethz.ch wrote: Therneau, Terry M , Ph D thern...@mayo.edu on Tue, 28 Oct 2014 07:44:20 -0500 writes: Martin, I can't imagine using such a function myself, the reason being that as Bill and Duncan point out, the correct answer depends on the situation. yes, of course. OTOH, if the function is used for argument checking inside another function, using such an is.whole(.) may come as a handy, and well readable {because self explaining} expression. But given the regular reappearance of this topic, I think that perhaps creation of your function is a good idea, largely to function as a repository for the knowlege. If one takes that view, then perhas the function has two optional arguments: case and tolerance. The first would choose a scenario of exact, numeric, count, etc, where exact refers to Duncan's case, numeric to your default, and count to Bill's a+1 a. The second argument would be rarely used. The primary point of the function would be the Details section of its manual page. Whenver the issue comes up the response could then be see the is.whole() function and its documentation. Terry T. Thank you, Duncan, and Terry, Yes, indeed, a primary point of the function would just be that: A coherent place to point to (and \link{.} to e.g. from the as.integer help page). Apropos optional arguments and their defaults: It may indeed be a better (than sfsmisc::is.whole 's default) idea to use a default tolerence = 0 rather than sqrt(.Machine$double.eps). .. and I think the argument / principle of thinking of what happens when integer - indexing with such numbers is also aa good one. That one has the drawback of asymmetry, i.e., of treating 4 + 1e-10 very differently than 4 - 1e-10 Martin On 10/28/2014 06:00 AM, r-devel-requ...@r-project.org wrote: Diverted to R-devel, as I'm requesting comments about a proposal to add is.whole() to R just so this issue does not trail on for centuries (;-), see below. __ R-devel@r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-devel -- Gregory (Greg) L. Snow Ph.D. 538...@gmail.com __ R-devel@r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-devel
Re: [Rd] OSX Yosemite (10.10): Are package binaries the same as for OSX Mavericks (10.9)?
On Mon, Oct 27, 2014 at 12:33 PM, Dan Tenenbaum dtene...@fredhutch.org wrote: - Original Message - From: Henrik Bengtsson h...@biostat.ucsf.edu To: Dan Tenenbaum dtene...@fredhutch.org Cc: R-devel r-devel@r-project.org Sent: Monday, October 27, 2014 12:21:49 PM Subject: Re: [Rd] OSX Yosemite (10.10): Are package binaries the same as for OSX Mavericks (10.9)? On Mon, Oct 27, 2014 at 11:23 AM, Dan Tenenbaum dtene...@fredhutch.org wrote: - Original Message - From: Dan Tenenbaum dtene...@fredhutch.org To: Henrik Bengtsson h...@biostat.ucsf.edu Cc: R-devel r-devel@r-project.org Sent: Monday, October 27, 2014 11:21:59 AM Subject: Re: [Rd] OSX Yosemite (10.10): Are package binaries the same as for OSX Mavericks (10.9)? - Original Message - From: Henrik Bengtsson h...@biostat.ucsf.edu To: R-devel r-devel@r-project.org Sent: Monday, October 27, 2014 11:16:10 AM Subject: [Rd] OSX Yosemite (10.10): Are package binaries the same as for OSX Mavericks (10.9)? I'm trying to help someone to troubleshoot possible OSX Yosemite issues, but I've only got access to OSX ( 10.9) so I cannot check myself. When building/installing binary R packages, there are different binaries depending on OSX version. For instance, CRAN provides different binaries for 'OS X Snow Leopard' and 'OS X Mavericks', e.g. http://cran.r-project.org/web/packages/matrixStats/index.html. What about the new OSX Yosemite? From http://cran.r-project.org/doc/manuals/r-devel/R-admin.html#Yosemite it looks like its binaries are the same/compatible with those of 'OS X Mavericks' - can someone please confirm this? Another way to put it, if a repository provides OSX Mavericks binaries will an OSX Yosemite user install these or we s/he fall back to installing from source? Yes, a Yosemite user will by default be installing packages built on Mavericks using the Mavericks build of R, and they should work. Provided of course that that Yosemite user is using the Mavericks build of R. They could also be using the Snow Leopard build of R which should also work, and would be installing by default packages build on Snow Leopard using the Snow Leopard build of R. Thanks for this Dan. As far as I understand, for an OSX user to install binary packages option 'pkgType' has to be set to either mac.binary or mac.binary.mavericks. A few questions for clarification: Q. Is it the default that 'pkgType' be set to mac.binary on OSX ( 10.9) and to mac.binary.mavericks on OSX (= 10.9)? Q. Are you saying that if an OSX (= 10.9) user uses options(pkgType=mac.binary), then install.packages() will install the OSX 10.6 (Snow Leopard) binaries *and* that these binaries are backward compatible and should work equally well? Q. In other words, if a user have problems with a particular OSX 10.9 (Mavericks) binary, would a first step of troubleshooting be to ask that user to try the OSX 10.6 (Snow Leopard) build? Q. If a user has options(pkgType=mac.binary.mavericks), but the repository does not provide such binaries, will install.packages() fall back to mac.binary, or will it go directly to source? First of all, this should be on R-SIG-Mac. I considered that, but I'm also asking this as a package developer and wonder what happens if someone installs my packages incorrectly and I need to troubleshoot what's reported as a bugs but may not be, so I though it would be more appropriate here. It all depends on what build of R you are using. You can be on Snow Leopard or later (including Mavericks and Yosemite) and use the Snow Leopard build. The default package type will be mac.binary. You can be on Mavericks or later and using the Mavericks build of R and your package type will by default be mac.binary.mavericks. Just for the record: I've verified that it is not possible to install the Mavericks build of R on a pre-Mavericks OSX version by mistake; on an OSX 10.6.8 machine I get: $ wget http://r.research.att.com/mavericks/R-3.1-branch/R-3.1-branch-mavericks.pkg $ sudo installer -pkg R-3.1-branch-mavericks.pkg -target / ... installer: This build of R requires Mac OS X 10.9 or higher. $ The two types of binary packages are NOT binary compatible! You should not mix and match them. (Technically, if a given package does not have native code in it, it should work, but you don't really want to go there.) I understand that packages without native code should work, but is there a reason for why R and install.packages() allows such mix and matching in the first place? I've tested install.packages(matrixStats, type=mac.binary.mavericks) on an OSX 10.6.8 machine and it install the package without complaints. Wouldn't it be better then if it gave an error: install.packages(matrixStats, type=mac.binary.mavericks) Installing package into '/Users/hb/Library/R/3.1/library' (as 'lib' is