R: selective import with importFrom: namespace issues [duplicate] - r

The "Writing R Extensions" manual provides the following guidance on when to use Imports or Depends:
The general rules are
Packages whose namespace only is needed to load the package using library(pkgname) must be listed in the ‘Imports’ field and not in the
‘Depends’ field.
Packages that need to be attached to successfully load the package using library(pkgname) must be listed in the ‘Depends’ field, only.
Can someone provide a bit more clarity on this? How do I know when my package only needs namespaces loaded versus when I need a package to be attached? What are examples of both? I think the typical package is just a collection of functions that sometimes call functions in other packages (where some bit of work has already been coded-up). Is this scenario 1 or 2 above?
Edit
I wrote a blog post with a section on this specific topic (search for 'Imports v Depends'). The visuals make it a lot easier to understand.

"Imports" is safer than "Depends" (and also makes a package using it a 'better citizen' with respect to other packages that do use "Depends").
A "Depends" directive attempts to ensure that a function from another package is available by attaching the other package to the main search path (i.e. the list of environments returned by search()). This strategy can, however, be thwarted if another package, loaded later, places an identically named function earlier on the search path. Chambers (in SoDA) uses the example of the function "gam", which is found in both the gam and mgcv packages. If two other packages were loaded, one of them depending on gam and one depending on mgcv, the function found by calls to gam() would depend on the order in which they those two packages were attached. Not good.
An "Imports" directive should be used for any supporting package whose functions are to be placed in <imports:packageName> (searched immediately after <namespace:packageName>), instead of on the regular search path. If either one of the packages in the example above used the "Imports" mechanism (which also requires import or importFrom directives in the NAMESPACE file), matters would be improved in two ways. (1) The package would itself gain control over which mgcv function is used. (2) By keeping the main search path clear of the imported objects, it would not even potentially break the other package's dependency on the other mgcv function.
This is why using namespaces is such a good practice, why it is now enforced by CRAN, and (in particular) why using "Imports" is safer than using "Depends".
Edited to add an important caveat:
There is one unfortunately common exception to the advice above: if your package relies on a package A which itself "Depends" on another package B, your package will likely need to attach A with a "Depends directive.
This is because the functions in package A were written with the expectation that package B and its functions would be attached to the search() path.
A "Depends" directive will load and attach package A, at which point package A's own "Depends" directive will, in a chain reaction, cause package B to be loaded and attached as well. Functions in package A will then be able to find the functions in package B on which they rely.
An "Imports" directive will load but not attach package A and will neither load nor attach package B. ("Imports", after all, expects that package writers are using the namespace mechanism, and that package A will be using "Imports" to point to any functions in B that it need access to.) Calls by your functions to any functions in package A which rely on functions in package B will consequently fail.
The only two solutions are to either:
Have your package attach package A using a "Depends" directive.
Better in the long run, contact the maintainer of package A and ask them to do a more careful job of constructing their namespace (in the words of Martin Morgan in this related answer).

Hadley Wickham gives an easy explanation (http://r-pkgs.had.co.nz/namespace.html):
Listing a package in either Depends or Imports ensures that it’s
installed when needed. The main difference is that where Imports just
loads the package, Depends attaches it. There are no other
differences. [...]
Unless there is a good reason otherwise, you should always list
packages in Imports not Depends. That’s because a good package is
self-contained, and minimises changes to the global environment
(including the search path). The only exception is if your package is
designed to be used in conjunction with another package. For example,
the analogue package builds on top of vegan. It’s not useful without
vegan, so it has vegan in Depends instead of Imports. Similarly,
ggplot2 should really Depend on scales, rather than Importing it.

Chambers in SfDA says to use 'Imports' when this package uses a 'namespace' mechanism and since all packages are now required to have them, then the answer might now be always use 'Imports'. In the past packages could have been loaded without actually having namespaces and in that case you would need to have used Depends.

Here is a simple question to help you decide which to use:
Does your package require the end user to have direct access to the functions of another package?
NO -> Imports (most common answer)
YES -> Depends
The only time you should use 'Depends' is when your package is an add-on or companion to another package, where your end user will be using functions from both your package and the 'Depends' package in their code. If your end user will only be interfacing with your functions, and the other package will only be doing work behind the scenes, then use 'Imports' instead.
The caveat to this is that if you add a package to 'Imports', as you usually should, your code will need to refer to functions from that package, using the full namespace syntax, e.g. dplyr::mutate(), instead of just mutate(). It makes the code a little clunkier to read, but it’s a small price to pay for better package hygiene.

Related

Run-time vs develop-time dependencies in R

I'm developing a package (golem) in R, and it returns a NOTE about excess package in an Import (DESCRIPTION):
checking package dependencies … NOTE
Imports includes 34 non-default packages.
Importing from so many packages makes the package vulnerable to any of
them becoming unavailable. Move as many as possible to Suggests and
use conditionally.
I have allocated some packages in Suggests (DESCRIPTION), like this:
usethis::use_package(package = "ggplot2", type = "Suggests")
usethis::use_package(package = "MASS", type = "Suggests")
I would like to know :
What is the difference between Imports (run-time) vs Suggests (develop-time) and if the latter has anything to do with the term "compile time" of other programming languages.
How do I know a package is needed by the user at runtime? Is there any universal rule for this (like a phrase to help you know)? And for Suggests?
In R, packages listed in the Imports clause of the DESCRIPTION file must be available or your package won't load. Normally they will all be loaded when your package is loaded, though it's possible to delay that by not importing anything, just using :: notation to access them.
Packages listed in the Suggests clause don't need to be available, and won't be automatically loaded. To access their functions, you normally call requireNamespace() to find out if the package is available, and if so use :: for access. If it is not available, your package should fail gracefully in whatever the user was trying to do, letting them know that they need to install the missing package if they want the task to succeed.
These aren't really "run-time" versus "develop-time" differences. It's all run-time.
There are two things in R that might be called "compile-time" in other languages. The best match is installing your package. That configures it to the particular R version and platform it is running on. R also has a "just-in-time" compiler that optimizes functions, but other than a bit of a speed increase that is pretty much invisible to the user.
I think #r2evans answered your second question clearly in a comment: the user needs a package to use functions that use that package. If some of your functions that use it are unlikely to be used by most users, use Suggests, and add the test.

Importing snowfall into custom R package

I'm developing an R package which needs to use parallelisation as made available by the snowfall package. snowfall doesn't seem to import the same was as other packages like ggplot2, data.table, etc. I've included snowfall, rlecuyer, and snow in the description file, name space file, and as an import argument in the function itself. When I try to access this function, I get the following error:
Error in sfInit() : could not find function "setDefaultClusterOptions"
The sfInit function seems to have a nostart / nostop argument which it says is related to nested usage of sfInit but that doesn't seem to do the trick for me either.
The actual code itself uses an sfInit (which is where I get the error), some sfExports and sfLibrarys, and an sfLapply.
Possible solution:
It seems to work if I move snow from the import section to the depends section in the Desciption file. I don't know why though.
When you include a package in 'Depends' when one attaches your package they also attach the package on which your package Depends to their namespace.
This and other differences between Depends and Imports is explained well in other questions on this site.
If you look at {snowfall}'s DESCRIPTION you'll see that it Depends on {snow}. It is plausible that the authors of snowfall know something we don't and that {snow} has to be attached to the global search path in order to work. In fact that is the top caveat in the top answer to the question I linked above...
... if your package relies on a package A which itself "Depends" on
another package B, your package will likely need to attach A with a
"Depends directive.
This is because the functions in package A were written with the
expectation that package B and its functions would be attached to the
search() path.
So, in your case, it just so happens that all {snowfall} wants is {snow} and you happened to provide it. However, it appears the more correct behavior may be for you to Depend on {snowfall} directly.
setDefaultClusterOptions is a function from the snow package. You need to import that too.

Use ::: or export everything when developing multiple, related R packages?

I am working on a family of R packages, all of which share substantial common code which is housed in an internal package, lets call it myPackageUtilities. So I have several packages
myPackage1, myPackage2, etc...
All of these packages depend on every method in myPackageUtilities. For a real-world example, please see statnet on CRAN. Each dependent package, of course, has
Depends: myPackageUtilities
in its DESCRIPTION file.
My question is: In the R code for myPackage1, which of the following two techniques for accessing methods from myPackageUtilities is preferable:
Use ::: to access the methods in myPackageUtilities, or
Export everything from myPackageUtilities (e.g. by including exportPattern("^[^\\.]") in the NAMESPACE)?
Option 2 clutters the end-user's search path, but the R gurus recommend against using :::.
Follow-up question: If (2) is the better choice, is there a way to export everything using roxygen2?
Suppose we have a package called randomUtils and this package has a function called sd that calculates the Slytherin Defiance Quotient.
Now I write a package called spellbound. If spellbound Depends on randomUtils, then randomUtils::sd will be found in the search path and can conflict with calculating standard deviation.
If spellbound Imports randomUtils, however, then R will install randomUtils but will not load it when spellbound is loaded. Thus, the new version of sd can't be found on the search path, but can still be accessed by randomUtils::sd
With an ever growing body of contributed work on CRAN, it is becoming very important that we use Imports as much as possible so that we don't introduce unexpected behaviors by having conflicting function definitions.
An example of when I have used Depends: when writing the HydeNet package, I wanted the end user to be able to use the rjags package in concert with HydeNet. So I put rjags in Depends so that library (HydeNet) would loaf both packages. (In other words, put rjags on the search path.
Moral of the story, if you don't intend for the user to directly access the functions, it should go in Imports.

I'm writing a package. How can make it such that when library(my_package) is called, other packages are loaded as well?

Title should be pretty clear I hope. I'm writing a package called forecasting, with imports for dplyr among other packages. With the imports written in to the DESCRIPTION file, I am able to force these other packages to be installed along with forecasting - is there an equivalent way to do this for the loading of the package? In other words, is there a way that when I load my package with library(forecasting), it automatically also loads dplyr and the other packages?
Thanks
Yes.
Re-read "Writing R Extensions". The Depends: forces both the initial installation as well as the loading of the depended-upon packages.
But these days you want Imports: along with importFrom() in the NAMESPACE file which is more fine-grained.
But first things first: get it working with Depends.
Edit:
Opps you're correct, the documentation I referenced is not a primary source. Perhaps this is better:
From the R documentation:
The ‘Depends’ field gives a comma-separated list of package names which this package depends on. Those packages will be attached before the current package when library or require is called.
and
The ‘Imports’ field lists packages whose namespaces are imported from (as specified in the NAMESPACE file) but which do not need to be attached. Namespaces accessed by the ‘::’ and ‘:::’ operators must be listed here, or in ‘Suggests’ or ‘Enhances’
Original:
From the R packages documentation:
Adding a package dependency here [the DESCRIPTION file] ensures that it’ll be installed. However, it does not mean that it will be attached along with your package (i.e., library(x)). The best practice is to explicitly refer to external functions using the syntax package::function(). This makes it very easy to identify which functions live outside of your package. This is especially useful when you read your code in the future.

Better explanation of when to use Imports/Depends

The "Writing R Extensions" manual provides the following guidance on when to use Imports or Depends:
The general rules are
Packages whose namespace only is needed to load the package using library(pkgname) must be listed in the ‘Imports’ field and not in the
‘Depends’ field.
Packages that need to be attached to successfully load the package using library(pkgname) must be listed in the ‘Depends’ field, only.
Can someone provide a bit more clarity on this? How do I know when my package only needs namespaces loaded versus when I need a package to be attached? What are examples of both? I think the typical package is just a collection of functions that sometimes call functions in other packages (where some bit of work has already been coded-up). Is this scenario 1 or 2 above?
Edit
I wrote a blog post with a section on this specific topic (search for 'Imports v Depends'). The visuals make it a lot easier to understand.
"Imports" is safer than "Depends" (and also makes a package using it a 'better citizen' with respect to other packages that do use "Depends").
A "Depends" directive attempts to ensure that a function from another package is available by attaching the other package to the main search path (i.e. the list of environments returned by search()). This strategy can, however, be thwarted if another package, loaded later, places an identically named function earlier on the search path. Chambers (in SoDA) uses the example of the function "gam", which is found in both the gam and mgcv packages. If two other packages were loaded, one of them depending on gam and one depending on mgcv, the function found by calls to gam() would depend on the order in which they those two packages were attached. Not good.
An "Imports" directive should be used for any supporting package whose functions are to be placed in <imports:packageName> (searched immediately after <namespace:packageName>), instead of on the regular search path. If either one of the packages in the example above used the "Imports" mechanism (which also requires import or importFrom directives in the NAMESPACE file), matters would be improved in two ways. (1) The package would itself gain control over which mgcv function is used. (2) By keeping the main search path clear of the imported objects, it would not even potentially break the other package's dependency on the other mgcv function.
This is why using namespaces is such a good practice, why it is now enforced by CRAN, and (in particular) why using "Imports" is safer than using "Depends".
Edited to add an important caveat:
There is one unfortunately common exception to the advice above: if your package relies on a package A which itself "Depends" on another package B, your package will likely need to attach A with a "Depends directive.
This is because the functions in package A were written with the expectation that package B and its functions would be attached to the search() path.
A "Depends" directive will load and attach package A, at which point package A's own "Depends" directive will, in a chain reaction, cause package B to be loaded and attached as well. Functions in package A will then be able to find the functions in package B on which they rely.
An "Imports" directive will load but not attach package A and will neither load nor attach package B. ("Imports", after all, expects that package writers are using the namespace mechanism, and that package A will be using "Imports" to point to any functions in B that it need access to.) Calls by your functions to any functions in package A which rely on functions in package B will consequently fail.
The only two solutions are to either:
Have your package attach package A using a "Depends" directive.
Better in the long run, contact the maintainer of package A and ask them to do a more careful job of constructing their namespace (in the words of Martin Morgan in this related answer).
Hadley Wickham gives an easy explanation (http://r-pkgs.had.co.nz/namespace.html):
Listing a package in either Depends or Imports ensures that it’s
installed when needed. The main difference is that where Imports just
loads the package, Depends attaches it. There are no other
differences. [...]
Unless there is a good reason otherwise, you should always list
packages in Imports not Depends. That’s because a good package is
self-contained, and minimises changes to the global environment
(including the search path). The only exception is if your package is
designed to be used in conjunction with another package. For example,
the analogue package builds on top of vegan. It’s not useful without
vegan, so it has vegan in Depends instead of Imports. Similarly,
ggplot2 should really Depend on scales, rather than Importing it.
Chambers in SfDA says to use 'Imports' when this package uses a 'namespace' mechanism and since all packages are now required to have them, then the answer might now be always use 'Imports'. In the past packages could have been loaded without actually having namespaces and in that case you would need to have used Depends.
Here is a simple question to help you decide which to use:
Does your package require the end user to have direct access to the functions of another package?
NO -> Imports (most common answer)
YES -> Depends
The only time you should use 'Depends' is when your package is an add-on or companion to another package, where your end user will be using functions from both your package and the 'Depends' package in their code. If your end user will only be interfacing with your functions, and the other package will only be doing work behind the scenes, then use 'Imports' instead.
The caveat to this is that if you add a package to 'Imports', as you usually should, your code will need to refer to functions from that package, using the full namespace syntax, e.g. dplyr::mutate(), instead of just mutate(). It makes the code a little clunkier to read, but it’s a small price to pay for better package hygiene.

Resources