Modules are a useful mechanism for providing and including names between different parts of a program that is split across files. One limitation of modules is that importing or including names from a module is strictly additive – including can only make more names available. For different course contexts and applications, sometimes we want more control over exactly which names are available, including leaving some out, or redefining some common names ourselves.
This is what use context is for. The meaning of use context is to start the program in an context with no names at all available, and then adding only what’s provided from the module listed after context <import>. The <import> part of use context line can use any of the dependency types.
As an example, consider an assignment where students write their own implementation of lists, defining functions named map, filter, and so on. In this case, the context they use should not be the default context (because it already has map and so on):
use context essentials2021 data List<A>: | empty | link(first :: A, rest :: List<A>) end # causes a shadowing error because empty and link already are defined and # provided by essentials2021
You could create a context with just a limited set of global names, and publish it as a shared-gdrive module:
# In file "list-assignment-context.arr" use context global # This is a good default environment to use when constructing namespaces # Basic functions/types like num-max, to-string, Number, String import global as G provide from G: *, type * end # Option and Either might be used as return types for some list functions, # so provide those import option as O provide from O: *, type * end import either as E provide from E: *, type * end
Then students could use that module as their context:
use context shared-gdrive("list-assignment-context.arr", "google-id-goes-here") data List<A>: | empty | link(first :: A, rest :: List<A>) end
By default, until use context was released, all Pyret programs had a single default context. It had provided common names related to, for example, Lists and Option. Now files that don’t have a use line are treated as if they start with use context essentials2020, which is equivalent to this historically available default context. The online environment code.pyret.org inserts use context essentialsYEAR into new programs, where YEAR will change when new useful names are available. Notably, essentials2021 includes all of the names for the untyped Image library by default, removing the need to include image in new programs.
Existing contexts provided by Pyret, like essentials2020 and essentials2021, aren’t intended to change the names they provide, so files with a use context line won’t have their set of available names changed with updates to the language. This is important for forwards compatibility, because Pyret is particular about shadowing, so context stability ensures that definitions in files using a context in this way won’t suddenly shadow a newly-provided name after an update.
In general, if you are the author of a context, you should carefully consider whether you want to add new names to an existing context, or create a new context with new names. The decision is mostly based on if active users are likely to have their own definitions of newly-provided names, because changing the context could cause a shadowing error in that case.
This means that sometimes programmers will want to manually update the year in a context, if it’s convenient for getting access to some new library function. Of course, they could always add the appropriate import statement to access it directly.