1. 注意
where <debug> $[some message [varname] ]
与
where <debug> [varname]
的区别。 如果varname代表一个数组/List，那么前者的输出为：
abc
后者的输出为
["a","b","c"]
见到前者的输出时，并不能由此就认定该变量没有保存一个list。
2. spoofax的api manual
http://releases.strategoxt.org/docs/api/libstratego-lib/stable/docs/html/index_all.html
3. 一些hints
Look-up transformations map names to information about the named element, e.g. a class name to a list of names of its ancestors. Since you have only a name, you typically have to get the declaration of the named element. Remember, you stored mappings from names to declarations earlier in this assignment. Either during or after this step, you project the declaration to the information you are interested in. In the ancestors example, you are not interested in the whole declaration but only in the name of the parent. You have already a projection strategy for this! You can use these strategies from the namespace library to get or project declarations:
- get-declarations(|namespace) maps a name to a list of its declarations in the given namespace.
- get-declarations(test | namespace) maps a name to a list of its declarations in the given namespace for which the strategy test is successful.
- get-declaration(test | namespace) maps a name to a the first declaration in the given namespace for which the strategy test is successful.
- project-declarations(projection | namespace) maps a name to a list of elements which are projections of its definitions in the given namespace.
- project-declaration(projection | namespace) maps a name to an element which is the first projection of one of its definitions in the given namespace.
- get-all-declarations(|namespace) maps any term to a list of all declarations in the given namespace.
- get-all-declarations(test | namespace) maps any term to a list of all declarations in the given namespace for which the strategy test is successful.
- project-all-declarations(projection | namespace) maps any term to a list of elements which are projections of the definitions in the given namespace.
Once you got the information you are interested in, you process this information further, maybe in another look-up or in a recursive call. Typical processes are mapping elements in a list to new elements in a new list or to filter elements from a list. Generic traversals are usually not involved here. The following strategies from the Stratego library might help you:
- elem succeeds if a term e is element of a list l. Needs to be applied to the tuple (e, l).
- fetch-elem(s) maps a list l to a single element e. The strategy tries to apply s to the elements of l. e will be the result of the first successful application.
- map(s) applies s to each element of a list. The results form a new list which is the result of the map.
- filter(s) tries to apply s to the elements of a list. Results of successful applications form a new list which is the result of the filter.
- where(s) tests if s can be applied to the current term without changing it.
- concat maps a list of lists to the concatenation of these lists.
- reverse maps a list to a reversed version of that list.
When you define recursive rewrite rules, start with the base cases. Define one rewrite rule for each base case. In the ancestors example, a base case is a class without a parent. Specify this condition in the where clause of the rewrite rule. Next, you can define recursive rewrite rules. You specify how to build the overall result from the result of a recursive call in the where clause. In the ancestors example, you need a rule for a class with a parent class. The ancestors of such a class include the ancestors of the parent class. You need to make sure that the recursion always terminate. In the ancestors example, you have to deal with cyclic inheritance. A possible solution is to add an additional parameter to your rewrite rules which accumulates the ancestor classes. You can then add another base case which checks if the current class is already in this accumulated list.
Note: If you are not familiar with the concept of an accumulator, this is a good point to change this. The inverse example from the lecture on Term Rewriting uses an accumulator.
