1 Class and Module Attribute Assignment
2 =====================================
3
4 Verify that the context information is correctly set, particularly for the unoptimised
5 cases.
6
7 Update docs/assignment.txt.
8
9 Dynamic Attribute Access
10 ========================
11
12 Consider explicit accessor initialisation.
13
14 Attribute Usage
15 ===============
16
17 Consider attribute assignment observations, along with the possibility of class attribute
18 assignment.
19
20 Note direct assignments as usual, indirect assignments via the attribute usage
21 mechanism. During attribute collection and inference, add assigned values to all
22 inferred targets.
23
24 Since class attributes can be assigned, StoreAttrIndex would no longer need to reject
25 static attributes, although this might still be necessary where attribute usage analysis
26 has not been performed.
27
28 Potentially consider changing static attribute details to use object-relative offsets in
29 order to simplify the instruction implementations. This might allow us to eliminate the
30 static attribute flag for attributes in the object table, at least at run-time.
31
32 Consider attribute usage observations being suspended inside blocks where AttributeError
33 may be caught (although this doesn't anticipate such exceptions being caught outside a
34 function altogether).
35
36 Consider type deduction and its consequences where types belong to the same hierarchy
37 and where a guard could be generated for the most general type.
38
39 Consider permitting multiple class alternatives where the attributes are all identical.
40
41 Support class attribute positioning similar to instance attribute positioning, potentially
42 (for both) based on usage observations. For example, if __iter__ is used on two classes,
43 the class attribute could be exposed at a similar relative position to the class (and
44 potentially accessible using a LoadAttr-style instruction).
45
46 **** Constant attribute users need not maintain usage since they are already resolved. ****
47
48 Loop entry points should capture usage to update later assignments in the loop.
49 The continue and break statements should affect usage propagation.
50
51 Consider handling CallFunc in micropython.inspect in order to produce instances of specific classes.
52 Then, consider adding support for guard removal/verification where known instances are involved.
53 Consider handling branches of values within namespaces in order to support more precise value usage.
54
55 Frame Optimisations
56 ===================
57
58 Stack frame replacement where a local frame is unused after a call, such as in a tail call
59 situation.
60
61 Local assignment detection plus frame re-use. Example: slice.__init__ calls
62 xrange.__init__ with the same arguments which are unchanged in xrange.__init__. There is
63 therefore no need to build a new frame for this call.
64
65 Function Specialisation
66 =======================
67
68 Specialisation of certain functions, such as isinstance(x, cls) where cls is a known
69 constant.
70
71 Structure and Object Table Optimisations
72 ========================================
73
74 Fix object table entries for attributes not provided by any known object, or provide an
75 error, potentially overridden by options. For example, the augmented assignment methods
76 are not supported by the built-in objects and thus the operator module functions cause
77 the compilation to fail. Alternatively, just supply the methods since something has to do
78 so in the builtins.
79
80 Consider attribute merging where many attributes are just aliases for the same underlying
81 definition.
82
83 Consider references to defaults as occurring only within the context of a particular
84 function, thus eliminating default value classes if such functions are not themselves
85 invoked.
86
87 Scope Handling
88 ==============
89
90 Consider merging the InspectedModule.store tests with the scope conflict handling.
91
92 Consider labelling _scope on assignments and dealing with the assignment of removed
93 attributes, possibly removing the entire assignment, and distinguishing between such cases
94 and unknown names.
95
96 Check name origin where multiple branches could yield multiple scope interpretations:
97
98 ----
99 try:
100 set # built-in name
101 except NameError:
102 from sets import Set as set # local definition of name
103
104 set # could be confused by the local definition at run-time
105 ----
106
107 Object Coverage
108 ===============
109
110 Support __init__ traversal (and other implicit names) more effectively.
111
112 Other
113 =====
114
115 Support tuple as a function returning any input tuple uncopied.
116
117 Check context_value initialisation (avoiding or handling None effectively).
118
119 __getitem__ could be written in Python, using a native method only to access fragments.
120
121 Consider better "macro" support where new expressions need to be generated and processed.
122
123 Detect TestIdentity results involving constants, potentially optimising status-affected
124 instructions:
125
126 TestIdentity(x, y) # where x is always y
127 JumpIfFalse(...) # would be removed (never false)
128 JumpIfTrue(...) # changed to Jump(...)
129
130 Status-affected blocks could be optimised away for such constant-related results.
