paul@15 | 1 | Namespace Definition
|
paul@15 | 2 | ====================
|
paul@15 | 3 |
|
paul@15 | 4 | Module attributes are defined either at the module level or by global
|
paul@15 | 5 | statements.
|
paul@15 | 6 |
|
paul@15 | 7 | Class attributes are defined only within class statements.
|
paul@15 | 8 |
|
paul@15 | 9 | Instance attributes are defined only by assignments to attributes of self
|
paul@15 | 10 | within __init__ methods.
|
paul@15 | 11 |
|
paul@42 | 12 | Potential Restrictions
|
paul@42 | 13 | ----------------------
|
paul@42 | 14 |
|
paul@42 | 15 | Names of classes and functions could be restricted to only refer to those
|
paul@42 | 16 | objects within the same namespace. If redefinition were to occur, or if
|
paul@42 | 17 | multiple possibilities were present, these restrictions could be moderated as
|
paul@42 | 18 | follows:
|
paul@42 | 19 |
|
paul@42 | 20 | * Classes assigned to the same name could provide the union of their
|
paul@42 | 21 | attributes. This would, however, cause a potential collision of attribute
|
paul@42 | 22 | definitions such as methods.
|
paul@42 | 23 |
|
paul@42 | 24 | * Functions, if they share compatible signatures, could share parameter list
|
paul@42 | 25 | definitions.
|
paul@42 | 26 |
|
paul@11 | 27 | Data Structures
|
paul@11 | 28 | ===============
|
paul@11 | 29 |
|
paul@45 | 30 | The fundamental "value type" is a pair of references: one pointing to the
|
paul@45 | 31 | referenced object represented by the interchangeable value; one referring to
|
paul@45 | 32 | the context of the referenced object, typically the object through which the
|
paul@45 | 33 | referenced object was acquired as an attribute.A
|
paul@45 | 34 |
|
paul@45 | 35 | Value Layout
|
paul@45 | 36 | ------------
|
paul@45 | 37 |
|
paul@45 | 38 | 0 1
|
paul@45 | 39 | object context
|
paul@45 | 40 | reference reference
|
paul@45 | 41 |
|
paul@52 | 42 | Acquiring Values
|
paul@52 | 43 | ----------------
|
paul@52 | 44 |
|
paul@52 | 45 | Values are acquired through name lookups and attribute access, yielding
|
paul@52 | 46 | the appropriate object reference together with a context reference as
|
paul@52 | 47 | indicated in the following table:
|
paul@52 | 48 |
|
paul@52 | 49 | Type of Access Context Notes
|
paul@52 | 50 | -------------- ------- -----
|
paul@52 | 51 |
|
paul@52 | 52 | Local name Preserved Functions provide no context
|
paul@52 | 53 |
|
paul@52 | 54 | Global name Preserved Modules provide no context
|
paul@52 | 55 |
|
paul@52 | 56 | Class-originating Accessor Methods acquire the context of their
|
paul@52 | 57 | attribute -or- accessor if an instance...
|
paul@52 | 58 | Preserved or retain the original context if the
|
paul@52 | 59 | accessor is a class
|
paul@52 | 60 |
|
paul@52 | 61 | Instance-originating Preserved Methods retain their original context
|
paul@52 | 62 | attribute
|
paul@52 | 63 |
|
paul@53 | 64 | There may be some scope for simplifying the above, to the detriment of Python
|
paul@52 | 65 | compatibility, since the unbound vs. bound methods situation can be confusing.
|
paul@52 | 66 |
|
paul@45 | 67 | Objects
|
paul@45 | 68 | -------
|
paul@45 | 69 |
|
paul@11 | 70 | Since classes, functions and instances are all "objects", each must support
|
paul@11 | 71 | certain features and operations in the same way.
|
paul@11 | 72 |
|
paul@11 | 73 | The __class__ Attribute
|
paul@11 | 74 | -----------------------
|
paul@11 | 75 |
|
paul@11 | 76 | All objects support the __class__ attribute:
|
paul@11 | 77 |
|
paul@11 | 78 | Class: refers to the type class (type.__class__ also refers to the type class)
|
paul@11 | 79 | Function: refers to the function class
|
paul@11 | 80 | Instance: refers to the class instantiated to make the object
|
paul@11 | 81 |
|
paul@11 | 82 | Invocation
|
paul@11 | 83 | ----------
|
paul@11 | 84 |
|
paul@11 | 85 | The following actions need to be supported:
|
paul@11 | 86 |
|
paul@11 | 87 | Class: create instance, call __init__ with instance, return object
|
paul@11 | 88 | Function: call function body, return result
|
paul@11 | 89 | Instance: call __call__ method, return result
|
paul@11 | 90 |
|
paul@11 | 91 | Structure Layout
|
paul@11 | 92 | ----------------
|
paul@11 | 93 |
|
paul@11 | 94 | A suitable structure layout might be something like this:
|
paul@11 | 95 |
|
paul@11 | 96 | 0 1 2 3 4
|
paul@11 | 97 | classcode invocation __class__ attribute ...
|
paul@11 | 98 | reference reference reference
|
paul@11 | 99 |
|
paul@11 | 100 | Here, the classcode refers to the attribute lookup table for the object. Since
|
paul@11 | 101 | classes and instances share the same classcode, they might resemble the
|
paul@11 | 102 | following:
|
paul@11 | 103 |
|
paul@11 | 104 | Class C:
|
paul@11 | 105 |
|
paul@11 | 106 | 0 1 2 3 4
|
paul@11 | 107 | code for C __new__ class type attribute ...
|
paul@11 | 108 | reference reference reference
|
paul@11 | 109 |
|
paul@11 | 110 | Instance of C:
|
paul@11 | 111 |
|
paul@11 | 112 | 0 1 2 3 4
|
paul@11 | 113 | code for C C.__call__ class C attribute ...
|
paul@11 | 114 | reference reference reference
|
paul@11 | 115 | (if exists)
|
paul@11 | 116 |
|
paul@11 | 117 | The __new__ reference would lead to code consisting of the following
|
paul@11 | 118 | instructions:
|
paul@11 | 119 |
|
paul@11 | 120 | create instance for C
|
paul@11 | 121 | call C.__init__(instance, ...)
|
paul@11 | 122 | return instance
|
paul@11 | 123 |
|
paul@11 | 124 | If C has a __call__ attribute, the invocation "slot" of C instances would
|
paul@11 | 125 | refer to the same thing as C.__call__.
|
paul@11 | 126 |
|
paul@11 | 127 | For functions, the same general layout applies:
|
paul@11 | 128 |
|
paul@11 | 129 | Function f:
|
paul@11 | 130 |
|
paul@11 | 131 | 0 1 2 3 4
|
paul@11 | 132 | code for code class attribute ...
|
paul@11 | 133 | function reference function reference
|
paul@11 | 134 | reference
|
paul@11 | 135 |
|
paul@37 | 136 | Here, the code reference would lead to code for the function. Note that the
|
paul@37 | 137 | function locals are completely distinct from this structure and are not
|
paul@37 | 138 | comparable to attributes.
|
paul@37 | 139 |
|
paul@38 | 140 | For modules, there is no meaningful invocation reference:
|
paul@37 | 141 |
|
paul@37 | 142 | Module m:
|
paul@37 | 143 |
|
paul@37 | 144 | 0 1 2 3 4
|
paul@38 | 145 | code for m (unused) module type attribute ...
|
paul@38 | 146 | reference (global)
|
paul@37 | 147 | reference
|
paul@11 | 148 |
|
paul@38 | 149 | Both classes and modules have code in their definitions, but this would be
|
paul@38 | 150 | generated in order and not referenced externally.
|
paul@38 | 151 |
|
paul@11 | 152 | Invocation Operation
|
paul@11 | 153 | --------------------
|
paul@11 | 154 |
|
paul@11 | 155 | Consequently, regardless of the object an invocation is always done as
|
paul@11 | 156 | follows:
|
paul@11 | 157 |
|
paul@11 | 158 | get invocation reference (at object+1)
|
paul@11 | 159 | jump to reference
|
paul@11 | 160 |
|
paul@11 | 161 | Additional preparation is necessary before the above code: positional
|
paul@11 | 162 | arguments must be saved to the parameter stack, and keyword arguments must be
|
paul@11 | 163 | resolved and saved to the appropriate position in the parameter stack.
|
paul@11 | 164 |
|
paul@11 | 165 | Attribute Operations
|
paul@11 | 166 | --------------------
|
paul@11 | 167 |
|
paul@53 | 168 | Attribute access needs to go through the attribute lookup table. Some
|
paul@53 | 169 | optimisations are possible and are described in the appropriate section.
|
paul@53 | 170 |
|
paul@53 | 171 | One important aspect of attribute access is the appropriate setting of the
|
paul@53 | 172 | context in the acquired attribute value. From the table describing the
|
paul@53 | 173 | acquisition of values, it is clear that the principal exception is that where
|
paul@53 | 174 | a class-originating attribute is accessed on an instance. Consequently, the
|
paul@53 | 175 | following algorithm could be employed once an attribute has been located:
|
paul@53 | 176 |
|
paul@53 | 177 | 1. If the attribute's context is a special value, indicating that it should
|
paul@53 | 178 | be replaced upon instance access, then proceed to the next step;
|
paul@53 | 179 | otherwise, acquire both the context and the object as they are.
|
paul@53 | 180 |
|
paul@53 | 181 | 2. If the accessor is an instance, use that as the value's context, acquiring
|
paul@53 | 182 | only the object from the attribute.
|
paul@53 | 183 |
|
paul@53 | 184 | Where accesses can be determined ahead of time (as discussed in the
|
paul@53 | 185 | optimisations section), the above algorithm may not necessarily be employed in
|
paul@53 | 186 | the generated code for some accesses.
|
paul@21 | 187 |
|
paul@21 | 188 | Instruction Evaluation Model
|
paul@21 | 189 | ============================
|
paul@21 | 190 |
|
paul@21 | 191 | Programs use a value stack where evaluated instructions may save their
|
paul@21 | 192 | results. A value stack pointer indicates the top of this stack. In addition, a
|
paul@21 | 193 | separate stack is used to record the invocation frames. All stack pointers
|
paul@21 | 194 | refer to the next address to be used by the stack, not the address of the
|
paul@21 | 195 | uppermost element.
|
paul@21 | 196 |
|
paul@21 | 197 | Frame Stack Value Stack
|
paul@21 | 198 | ----------- ----------- Address of Callable
|
paul@21 | 199 | -------------------
|
paul@21 | 200 | previous ...
|
paul@21 | 201 | current ------> callable -----> identifier
|
paul@21 | 202 | arg1 reference to code
|
paul@21 | 203 | arg2
|
paul@21 | 204 | arg3
|
paul@21 | 205 | local4
|
paul@21 | 206 | local5
|
paul@21 | 207 | ...
|
paul@21 | 208 |
|
paul@21 | 209 | Loading local names is a matter of performing frame-relative accesses to the
|
paul@21 | 210 | value stack.
|
paul@21 | 211 |
|
paul@21 | 212 | Invocations and Argument Evaluation
|
paul@21 | 213 | -----------------------------------
|
paul@21 | 214 |
|
paul@21 | 215 | When preparing for an invocation, the caller first sets the invocation frame
|
paul@21 | 216 | pointer. Then, positional arguments are added to the stack such that the first
|
paul@21 | 217 | argument positions are filled. A number of stack locations for the remaining
|
paul@21 | 218 | arguments specified in the program are then reserved. The names of keyword
|
paul@21 | 219 | arguments are used (in the form of table index values) to consult the
|
paul@21 | 220 | parameter table and to find the location in which such arguments are to be
|
paul@21 | 221 | stored.
|
paul@21 | 222 |
|
paul@21 | 223 | fn(a, b, d=1, e=2, c=3) -> fn(a, b, c, d, e)
|
paul@21 | 224 |
|
paul@21 | 225 | Value Stack
|
paul@21 | 226 | -----------
|
paul@21 | 227 |
|
paul@21 | 228 | ... ... ... ...
|
paul@21 | 229 | fn fn fn fn
|
paul@21 | 230 | a a a a
|
paul@21 | 231 | b b b b
|
paul@21 | 232 | ___ ___ ___ --> 3
|
paul@21 | 233 | ___ --> 1 1 | 1
|
paul@21 | 234 | ___ | ___ --> 2 | 2
|
paul@21 | 235 | 1 ----------- 2 ----------- 3 -----------
|
paul@21 | 236 |
|
paul@21 | 237 | Conceptually, the frame can be considered as a collection of attributes, as
|
paul@21 | 238 | seen in other kinds of structures:
|
paul@21 | 239 |
|
paul@21 | 240 | Frame for invocation of fn:
|
paul@21 | 241 |
|
paul@21 | 242 | 0 1 2 3 4 5
|
paul@21 | 243 | code a b c d e
|
paul@21 | 244 | reference
|
paul@21 | 245 |
|
paul@21 | 246 | However, where arguments are specified positionally, such "attributes" are not
|
paul@21 | 247 | set using a comparable approach to that employed with other structures.
|
paul@21 | 248 | Keyword arguments are set using an attribute-like mechanism, though, where the
|
paul@21 | 249 | position of each argument discovered using the parameter table.
|
paul@21 | 250 |
|
paul@45 | 251 | Method invocations incorporate an implicit first argument which is obtained
|
paul@45 | 252 | from the context of the method:
|
paul@45 | 253 |
|
paul@45 | 254 | method(a, b, d=1, e=2, c=3) -> method(self, a, b, c, d, e)
|
paul@45 | 255 |
|
paul@45 | 256 | Value Stack
|
paul@45 | 257 | -----------
|
paul@45 | 258 |
|
paul@45 | 259 | ...
|
paul@45 | 260 | method
|
paul@45 | 261 | context of method
|
paul@45 | 262 | a
|
paul@45 | 263 | b
|
paul@45 | 264 | 3
|
paul@45 | 265 | 1
|
paul@45 | 266 | 2
|
paul@45 | 267 |
|
paul@45 | 268 | Although it could be possible to permit any object to be provided as the first
|
paul@45 | 269 | argument, in order to optimise instance attribute access in methods, we should
|
paul@45 | 270 | seek to restrict the object type.
|
paul@45 | 271 |
|
paul@21 | 272 | Tuples, Frames and Allocation
|
paul@21 | 273 | -----------------------------
|
paul@21 | 274 |
|
paul@21 | 275 | Using the approach where arguments are treated like attributes in some kind of
|
paul@21 | 276 | structure, we could choose to allocate frames in places other than a stack.
|
paul@21 | 277 | This would produce something somewhat similar to a plain tuple object.
|
paul@23 | 278 |
|
paul@23 | 279 | Optimisations
|
paul@23 | 280 | =============
|
paul@23 | 281 |
|
paul@29 | 282 | Some optimisations around constant objects might be possible; these depend on
|
paul@29 | 283 | the following:
|
paul@29 | 284 |
|
paul@29 | 285 | * Reliable tracking of assignments: where assignment operations occur, the
|
paul@29 | 286 | target of the assignment should be determined if any hope of optimisation
|
paul@29 | 287 | is to be maintained. Where no guarantees can be made about the target of
|
paul@29 | 288 | an assignment, no assignment-related information should be written to
|
paul@29 | 289 | potential targets.
|
paul@29 | 290 |
|
paul@29 | 291 | * Objects acting as "containers" of attributes must be regarded as "safe":
|
paul@29 | 292 | where assignments are recorded as occurring on an attribute, it must be
|
paul@29 | 293 | guaranteed that no other unforeseen ways exist to assign to such
|
paul@29 | 294 | attributes.
|
paul@29 | 295 |
|
paul@29 | 296 | The discussion below presents certain rules which must be imposed to uphold
|
paul@29 | 297 | the above requirements.
|
paul@29 | 298 |
|
paul@30 | 299 | Safe Containers
|
paul@30 | 300 | ---------------
|
paul@28 | 301 |
|
paul@23 | 302 | Where attributes of modules, classes and instances are only set once and are
|
paul@23 | 303 | effectively constant, it should be possible to circumvent the attribute lookup
|
paul@28 | 304 | mechanism and to directly reference the attribute value. This technique may
|
paul@30 | 305 | only be considered applicable for the following "container" objects, subject
|
paul@30 | 306 | to the noted constraints:
|
paul@28 | 307 |
|
paul@30 | 308 | 1. For modules, "safety" is enforced by ensuring that assignments to module
|
paul@30 | 309 | attributes are only permitted within the module itself either at the
|
paul@30 | 310 | top-level or via names declared as globals. Thus, the following would not
|
paul@30 | 311 | be permitted:
|
paul@28 | 312 |
|
paul@28 | 313 | another_module.this_module.attr = value
|
paul@28 | 314 |
|
paul@29 | 315 | In the above, this_module is a reference to the current module.
|
paul@28 | 316 |
|
paul@30 | 317 | 2. For classes, "safety" is enforced by ensuring that assignments to class
|
paul@30 | 318 | attributes are only permitted within the class definition, outside
|
paul@30 | 319 | methods. This would mean that classes would be "sealed" at definition time
|
paul@30 | 320 | (like functions).
|
paul@28 | 321 |
|
paul@28 | 322 | Unlike the property of function locals that they may only sensibly be accessed
|
paul@28 | 323 | within the function in which they reside, these cases demand additional
|
paul@28 | 324 | controls or assumptions on or about access to the stored data. Meanwhile, it
|
paul@28 | 325 | would be difficult to detect eligible attributes on arbitrary instances due to
|
paul@28 | 326 | the need for some kind of type inference or abstract execution.
|
paul@28 | 327 |
|
paul@30 | 328 | Constant Attributes
|
paul@30 | 329 | -------------------
|
paul@30 | 330 |
|
paul@30 | 331 | When accessed via "safe containers", as described above, any attribute with
|
paul@30 | 332 | only one recorded assignment on it can be considered a constant attribute and
|
paul@30 | 333 | this eligible for optimisation, the consequence of which would be the
|
paul@30 | 334 | replacement of a LoadAttrIndex instruction (which needs to look up an
|
paul@30 | 335 | attribute using the run-time details of the "container" and the compile-time
|
paul@30 | 336 | details of the attribute) with a LoadAttr instruction.
|
paul@30 | 337 |
|
paul@30 | 338 | However, some restrictions exist on assignment operations which may be
|
paul@30 | 339 | regarded to cause only one assignment in the lifetime of a program:
|
paul@30 | 340 |
|
paul@30 | 341 | 1. For module attributes, only assignments at the top-level outside loop
|
paul@30 | 342 | statements can be reliably assumed to cause only a single assignment.
|
paul@30 | 343 |
|
paul@30 | 344 | 2. For class attributes, only assignments at the top-level within class
|
paul@30 | 345 | definitions and outside loop statements can be reliably assumed to cause
|
paul@30 | 346 | only a single assignment.
|
paul@30 | 347 |
|
paul@30 | 348 | All assignments satisfying the "safe container" requirements, but not the
|
paul@30 | 349 | requirements immediately above, should each be recorded as causing at least
|
paul@30 | 350 | one assignment.
|
paul@28 | 351 |
|
paul@29 | 352 | Additional Controls
|
paul@29 | 353 | -------------------
|
paul@29 | 354 |
|
paul@29 | 355 | For the above cases for "container" objects, the following controls would need
|
paul@29 | 356 | to apply:
|
paul@29 | 357 |
|
paul@29 | 358 | 1. Modules would need to be immutable after initialisation. However, during
|
paul@29 | 359 | initialisation, there remains a possibility of another module attempting
|
paul@29 | 360 | to access the original module. For example, if ppp/__init__.py contained
|
paul@29 | 361 | the following...
|
paul@29 | 362 |
|
paul@29 | 363 | x = 1
|
paul@29 | 364 | import ppp.qqq
|
paul@29 | 365 | print x
|
paul@29 | 366 |
|
paul@29 | 367 | ...and if ppp/qqq.py contained the following...
|
paul@29 | 368 |
|
paul@29 | 369 | import ppp
|
paul@29 | 370 | ppp.x = 2
|
paul@29 | 371 |
|
paul@29 | 372 | ...then the value 2 would be printed. Since modules are objects which are
|
paul@29 | 373 | registered globally in a program, it would be possible to set attributes
|
paul@29 | 374 | in the above way.
|
paul@29 | 375 |
|
paul@29 | 376 | 2. Classes would need to be immutable after initialisation. However, since
|
paul@29 | 377 | classes are objects, any reference to a class after initialisation could
|
paul@29 | 378 | be used to set attributes on the class.
|
paul@29 | 379 |
|
paul@29 | 380 | Solutions:
|
paul@29 | 381 |
|
paul@29 | 382 | 1. Insist on global scope for module attribute assignments.
|
paul@29 | 383 |
|
paul@29 | 384 | 2. Insist on local scope within classes.
|
paul@29 | 385 |
|
paul@29 | 386 | Both of the above measures need to be enforced at run-time, since an arbitrary
|
paul@29 | 387 | attribute assignment could be attempted on any kind of object, yet to uphold
|
paul@29 | 388 | the properties of "safe containers", attempts to change attributes of such
|
paul@29 | 389 | objects should be denied. Since foreseen attribute assignment operations have
|
paul@29 | 390 | certain properties detectable at compile-time, it could be appropriate to
|
paul@29 | 391 | generate special instructions (or modified instructions) during the
|
paul@29 | 392 | initialisation of modules and classes for such foreseen assignments, whilst
|
paul@29 | 393 | employing normal attribute assignment operations in all other cases. Indeed,
|
paul@29 | 394 | the StoreAttr instruction, which is used to set attributes in "safe
|
paul@29 | 395 | containers" would be used exclusively for this purpose; the StoreAttrIndex
|
paul@29 | 396 | instruction would be used exclusively for all other attribute assignments.
|
paul@29 | 397 |
|
paul@43 | 398 | To ensure the "sealing" of modules and classes, entries in the attribute
|
paul@43 | 399 | lookup table would encode whether a class or module is being accessed, so
|
paul@43 | 400 | that the StoreAttrIndex instruction could reject such accesses.
|
paul@43 | 401 |
|
paul@28 | 402 | Constant Attribute Values
|
paul@28 | 403 | -------------------------
|
paul@28 | 404 |
|
paul@29 | 405 | Where an attribute value is itself regarded as constant, is a "safe container"
|
paul@29 | 406 | and is used in an operation accessing its own attributes, the value can be
|
paul@29 | 407 | directly inspected for optimisations or employed in the generated code. For
|
paul@29 | 408 | the attribute values themselves, only objects of a constant nature may be
|
paul@28 | 409 | considered suitable for this particular optimisation:
|
paul@28 | 410 |
|
paul@28 | 411 | * Classes
|
paul@28 | 412 | * Modules
|
paul@28 | 413 | * Instances defined as constant literals
|
paul@28 | 414 |
|
paul@28 | 415 | This is because arbitrary objects (such as most instances) have no
|
paul@28 | 416 | well-defined form before run-time and cannot be investigated further at
|
paul@28 | 417 | compile-time or have a representation inserted into the generated code.
|
paul@29 | 418 |
|
paul@29 | 419 | Class Attributes and Access via Instances
|
paul@29 | 420 | -----------------------------------------
|
paul@29 | 421 |
|
paul@29 | 422 | Unlike module attributes, class attributes can be accessed in a number of
|
paul@29 | 423 | different ways:
|
paul@29 | 424 |
|
paul@29 | 425 | * Using the class itself:
|
paul@29 | 426 |
|
paul@29 | 427 | C.x = 123
|
paul@29 | 428 | cls = C; cls.x = 234
|
paul@29 | 429 |
|
paul@29 | 430 | * Using a subclass of the class (for reading attributes):
|
paul@29 | 431 |
|
paul@29 | 432 | class D(C):
|
paul@29 | 433 | pass
|
paul@29 | 434 | D.x # setting D.x would populate D, not C
|
paul@29 | 435 |
|
paul@29 | 436 | * Using instances of the class or a subclass of the class (for reading
|
paul@29 | 437 | attributes):
|
paul@29 | 438 |
|
paul@29 | 439 | c = C()
|
paul@29 | 440 | c.x # setting c.x would populate c, not C
|
paul@29 | 441 |
|
paul@29 | 442 | Since assignments are only achieved using direct references to the class, and
|
paul@29 | 443 | since class attributes should be defined only within the class initialisation
|
paul@29 | 444 | process, the properties of class attributes should be consistent with those
|
paul@29 | 445 | desired.
|
paul@29 | 446 |
|
paul@29 | 447 | Method Access via Instances
|
paul@29 | 448 | ---------------------------
|
paul@29 | 449 |
|
paul@29 | 450 | It is desirable to optimise method access, even though most method calls are
|
paul@29 | 451 | likely to occur via instances. It is possible, given the properties of methods
|
paul@29 | 452 | as class attributes to investigate the kind of instance that the self
|
paul@29 | 453 | parameter/local refers to within each method: it should be an instance either
|
paul@29 | 454 | of the class in which the method is defined or a compatible class, although
|
paul@29 | 455 | situations exist where this might not be the case:
|
paul@29 | 456 |
|
paul@29 | 457 | * Explicit invocation of a method:
|
paul@29 | 458 |
|
paul@29 | 459 | d = D() # D is not related to C
|
paul@29 | 460 | C.f(d) # calling f(self) in C
|
paul@29 | 461 |
|
paul@30 | 462 | If blatant usage of incompatible instances were somehow disallowed, it would
|
paul@30 | 463 | still be necessary to investigate the properties of an instance's class and
|
paul@30 | 464 | its relationship with other classes. Consider the following example:
|
paul@30 | 465 |
|
paul@30 | 466 | class A:
|
paul@30 | 467 | def f(self): ...
|
paul@30 | 468 |
|
paul@30 | 469 | class B:
|
paul@30 | 470 | def f(self): ...
|
paul@30 | 471 | def g(self):
|
paul@30 | 472 | self.f()
|
paul@30 | 473 |
|
paul@30 | 474 | class C(A, B):
|
paul@30 | 475 | pass
|
paul@30 | 476 |
|
paul@30 | 477 | Here, instances of B passed into the method B.g could be assumed to provide
|
paul@30 | 478 | access to B.f when self.f is resolved at compile-time. However, instances of C
|
paul@30 | 479 | passed into B.g would instead provide access to A.f when self.f is resolved at
|
paul@30 | 480 | compile-time (since the method resolution order is C, A, B instead of just B).
|
paul@30 | 481 |
|
paul@30 | 482 | One solution might be to specialise methods for each instance type, but this
|
paul@30 | 483 | could be costly. Another less ambitious solution might only involve the
|
paul@30 | 484 | optimisation of such internal method calls if an unambiguous target can be
|
paul@30 | 485 | resolved.
|
paul@30 | 486 |
|
paul@29 | 487 | Optimising Function Invocations
|
paul@29 | 488 | -------------------------------
|
paul@29 | 489 |
|
paul@29 | 490 | Where an attribute value is itself regarded as constant and is a function,
|
paul@29 | 491 | knowledge about the parameters of the function can be employed to optimise the
|
paul@29 | 492 | preparation of the invocation frame.
|