Constructive options for type 'environment'

Environments use reference semantics, they cannot be copied. An attempt to copy an environment would indeed yield a different environment and identical(env, copy) would be FALSE.
Moreover most environments have a parent (exceptions are emptyenv() and some rare cases where the parent is NULL) and thus to copy the environment we'd have to have a way to point to the parent, or copy it too.
For this reason environments are constructive's cryptonite. They make some objects impossible to reproduce exactly. And since every function or formula has one they're hard to avoid.

Usage

opts_environment(
  constructor = c(".env", "list2env", "as.environment", "new.env", "topenv",
    "new_environment", "predefine"),
  ...,
  recurse = FALSE
)

Arguments

constructor

String. Name of the function used to construct the environment, see Constructors section.

...

Additional options used by user defined constructors through the opts object

recurse

Boolean. Only considered if constructor is "list2env" or "new_environment". Whether to attempt to recreate all parent environments until a known environment is found, if FALSE (the default) we will use topenv() to find a known ancestor to set as the parent.

Value

An object of class <constructive_options/constructive_options_environment>

Details

In some case we can build code that points to a specific environment, namely:

• .GlobalEnv, .BaseNamespaceEnv, baseenv() and emptyenv() are used to construct the global environment, the base namespace, the base package environment and the empty environment

• Namespaces are constructed using asNamespace("pkg")

• Package environments are constructed using as.environment("package:pkg")

• "imports" environments are constructed with parent.env(asNamespace("pkg"))

• "lazydata" environments are constructed with getNamespaceInfo("pkg", "lazydata")

By default For other environments we use constructive's function constructive::.env(), it fetches the environment from its memory address and provides as additional information the sequence of parents until we reach a special environment (those enumerated above). The advantage of this approach is that it's readable and that the object is accurately reproduced. The inconvenient is that it's not stable between sessions. If an environment has a NULL parent it's always constructed with constructive::.env(), whatever the choice of the constructor.

Often however we wish to be able to reproduce from scratch a similar environment, so that we might run the constructed code later in a new session. We offer different different options to do this, with different trade-offs regarding accuracy and verbosity.

{constructive} will not signal any difference if it can reproduce an equivalent environment, defined as containing the same values and having a same or equivalent parent.

See also the ignore_function_env argument in ?compare_options, which disables the check of environments of function.

Constructors

We might set the constructor argument to:

• ".env" (default): use constructive::.env() to construct the environment from its memory address.

• "list2env": We construct the environment as a list then use base::list2env() to convert it to an environment and assign it a parent. By default we will use base::topenv() to construct a parent. If recurse is TRUE the parent will be built recursively so all ancestors will be created until we meet a known environment, this might be verbose and will fail if environments are nested too deep or have a circular relationship. If the environment is empty we use new.env(parent=) for a more economic syntax.

• "new_environment" : Similar to the above, but using rlang::new_environment().

• "new.env" : All environments will be recreated with the code "base::new.env()", without argument, effectively creating an empty environment child of the local (often global) environment. This is enough in cases where the environment doesn't matter (or matters as long as it inherits from the local environment), as is often the case with formulas. recurse is ignored.

• "as.environment" : we attempt to construct the environment as a list and use base::as.environment() on top of it, as in as.environment(list(a=1, b=2)), it will contain the same variables as the original environment but the parent will be the emptyenv(). recurse is ignored.

• "topenv" : we construct base::topenv(x), see ?topenv. recurse is ignored. This is the most accurate we can be when constructing only special environments.

• "predefine" : Building environments from scratch using the above methods can be verbose, sometimes redundant and sometimes even impossible due to circularity (e.g. an environment referencing itself). With "predefine" we define the environments and their content above the object returning call, using placeholder names ..env.1.., ..env.2.. etc. The caveat is that the created code won't be a single call and will create objects in the workspace. recurse is ignored.