mu/057immutable.cc

576 lines
18 KiB
C++
Raw Normal View History

//: Ingredients of a recipe are meant to be immutable unless they're also
//: products. This layer will start enforcing this check.
//:
//: One hole for now: variables in surrounding spaces are implicitly mutable.
//: [tag: todo]
:(scenario can_modify_ingredients_that_are_also_products)
# mutable container
def main [
local-scope
p:point <- merge 34, 35
p <- foo p
]
def foo p:point -> p:point [
local-scope
load-ingredients
p <- put p, x:offset, 34
]
$error: 0
:(scenario can_modify_ingredients_that_are_also_products_2)
def main [
local-scope
2016-09-17 19:55:10 +00:00
p:&:point <- new point:type
p <- foo p
]
# mutable address to container
2016-09-17 19:55:10 +00:00
def foo p:&:point -> p:&:point [
local-scope
load-ingredients
*p <- put *p, x:offset, 34
]
$error: 0
:(scenario can_modify_ingredients_that_are_also_products_3)
def main [
local-scope
2016-09-17 20:00:39 +00:00
p:&:@:num <- new number:type, 3
p <- foo p
]
# mutable address
2016-09-17 20:00:39 +00:00
def foo p:&:@:num -> p:&:@:num [
local-scope
load-ingredients
*p <- put-index *p, 0, 34
]
$error: 0
:(scenario ignore_literal_ingredients_for_immutability_checks)
def main [
local-scope
2016-09-17 19:55:10 +00:00
p:&:d1 <- new d1:type
2016-09-17 17:28:25 +00:00
q:num <- foo p
]
2016-09-17 19:55:10 +00:00
def foo p:&:d1 -> q:num [
local-scope
load-ingredients
2016-09-17 19:55:10 +00:00
x:&:d1 <- new d1:type
*x <- put *x, p:offset, 34 # ignore this 'p'
return 36
]
container d1 [
2016-09-17 17:28:25 +00:00
p:num
q:num
]
$error: 0
:(scenario cannot_modify_immutable_ingredients)
% Hide_errors = true;
def main [
local-scope
2016-09-17 19:55:10 +00:00
x:&:num <- new number:type
foo x
]
# immutable address to primitive
2016-09-17 19:55:10 +00:00
def foo x:&:num [
local-scope
load-ingredients
*x <- copy 34
]
+error: foo: cannot modify 'x' in instruction '*x <- copy 34' because it's an ingredient of recipe foo but not also a product
:(scenario cannot_modify_immutable_containers)
% Hide_errors = true;
def main [
local-scope
x:point-number <- merge 34, 35, 36
foo x
]
# immutable container
def foo x:point-number [
local-scope
load-ingredients
# copy an element: ok
y:point <- get x, xy:offset
# modify the element: boom
# This could be ok if y contains no addresses, but we're not going to try to be that smart.
# It also makes the rules easier to reason about. If it's just an ingredient, just don't try to change it.
y <- put y, x:offset, 37
]
+error: foo: cannot modify 'y' in instruction 'y <- put y, x:offset, 37' because that would modify 'x' which is an ingredient of recipe foo but not also a product
:(scenario can_modify_immutable_pointers)
def main [
local-scope
2016-09-17 19:55:10 +00:00
x:&:num <- new number:type
foo x
]
2016-09-17 19:55:10 +00:00
def foo x:&:num [
local-scope
load-ingredients
# modify the address, not the payload
x <- copy 0
]
$error: 0
:(scenario can_modify_immutable_pointers_but_not_their_payloads)
% Hide_errors = true;
def main [
local-scope
2016-09-17 19:55:10 +00:00
x:&:num <- new number:type
foo x
]
2016-09-17 19:55:10 +00:00
def foo x:&:num [
local-scope
load-ingredients
# modify address; ok
x <- new number:type
# modify payload: boom
# this could be ok, but we're not going to try to be that smart
*x <- copy 34
]
+error: foo: cannot modify 'x' in instruction '*x <- copy 34' because it's an ingredient of recipe foo but not also a product
:(scenario cannot_call_mutating_recipes_on_immutable_ingredients)
% Hide_errors = true;
def main [
local-scope
2016-09-17 19:55:10 +00:00
p:&:point <- new point:type
foo p
]
2016-09-17 19:55:10 +00:00
def foo p:&:point [
local-scope
load-ingredients
bar p
]
2016-09-17 19:55:10 +00:00
def bar p:&:point -> p:&:point [
local-scope
load-ingredients
# p could be modified here, but it doesn't have to be, it's already marked
# mutable in the header
]
+error: foo: cannot modify 'p' in instruction 'bar p' because it's an ingredient of recipe foo but not also a product
:(scenario cannot_modify_copies_of_immutable_ingredients)
% Hide_errors = true;
def main [
local-scope
2016-09-17 19:55:10 +00:00
p:&:point <- new point:type
foo p
]
2016-09-17 19:55:10 +00:00
def foo p:&:point [
local-scope
load-ingredients
2016-09-17 19:55:10 +00:00
q:&:point <- copy p
*q <- put *q, x:offset, 34
]
+error: foo: cannot modify 'q' in instruction '*q <- put *q, x:offset, 34' because that would modify p which is an ingredient of recipe foo but not also a product
:(scenario can_modify_copies_of_mutable_ingredients)
def main [
local-scope
2016-09-17 19:55:10 +00:00
p:&:point <- new point:type
foo p
]
2016-09-17 19:55:10 +00:00
def foo p:&:point -> p:&:point [
local-scope
load-ingredients
2016-09-17 19:55:10 +00:00
q:&:point <- copy p
*q <- put *q, x:offset, 34
]
$error: 0
:(scenario cannot_modify_address_inside_immutable_ingredients)
% Hide_errors = true;
container foo [
2016-09-17 20:00:39 +00:00
x:&:@:num # contains an address
]
def main [
# don't run anything
]
2016-09-17 19:55:10 +00:00
def foo a:&:foo [
local-scope
load-ingredients
2016-09-17 20:00:39 +00:00
x:&:@:num <- get *a, x:offset # just a regular get of the container
*x <- put-index *x, 0, 34 # but then a put-index on the result
]
+error: foo: cannot modify 'x' in instruction '*x <- put-index *x, 0, 34' because that would modify a which is an ingredient of recipe foo but not also a product
:(scenario cannot_modify_address_inside_immutable_ingredients_2)
container foo [
2016-09-17 20:00:39 +00:00
x:&:@:num # contains an address
]
def main [
# don't run anything
]
2016-09-17 19:55:10 +00:00
def foo a:&:foo [
local-scope
load-ingredients
b:foo <- merge 0
# modify b, completely unrelated to immutable ingredient a
2016-09-17 20:00:39 +00:00
x:&:@:num <- get b, x:offset
*x <- put-index *x, 0, 34
]
$error: 0
:(scenario cannot_modify_address_inside_immutable_ingredients_3)
% Hide_errors = true;
def main [
# don't run anything
]
2016-09-17 20:00:39 +00:00
def foo a:&:@:&:num [
local-scope
load-ingredients
2016-09-17 19:55:10 +00:00
x:&:num <- index *a, 0 # just a regular index of the array
*x <- copy 34 # but then modify the result
]
+error: foo: cannot modify 'x' in instruction '*x <- copy 34' because that would modify a which is an ingredient of recipe foo but not also a product
:(scenario cannot_modify_address_inside_immutable_ingredients_4)
def main [
# don't run anything
]
2016-09-17 20:00:39 +00:00
def foo a:&:@:&:num [
local-scope
load-ingredients
2016-09-17 20:00:39 +00:00
b:&:@:&:num <- new {(address number): type}, 3
# modify b, completely unrelated to immutable ingredient a
2016-09-17 19:55:10 +00:00
x:&:num <- index *b, 0
*x <- copy 34
]
$error: 0
:(scenario latter_ingredient_of_index_is_immutable)
def main [
# don't run anything
]
def foo a:&:@:&:@:num, b:num -> a:&:@:&:@:num [
local-scope
load-ingredients
x:&:@:num <- index *a, b
*x <- put-index *x, 0, 34
]
$error: 0
:(scenario can_traverse_immutable_ingredients)
container test-list [
2016-09-17 19:55:10 +00:00
next:&:test-list
]
def main [
local-scope
2016-09-17 19:55:10 +00:00
p:&:test-list <- new test-list:type
foo p
]
2016-09-17 19:55:10 +00:00
def foo p:&:test-list [
local-scope
load-ingredients
2016-09-17 19:55:10 +00:00
p2:&:test-list <- bar p
]
2016-09-17 19:55:10 +00:00
def bar x:&:test-list -> y:&:test-list [
local-scope
load-ingredients
y <- get *x, next:offset
]
$error: 0
:(scenario treat_optional_ingredients_as_mutable)
def main [
2016-09-17 19:55:10 +00:00
k:&:num <- new number:type
test k
]
# recipe taking an immutable address ingredient
2016-09-17 19:55:10 +00:00
def test k:&:num [
local-scope
load-ingredients
foo k
]
# ..calling a recipe with an optional address ingredient
def foo -> [
local-scope
load-ingredients
2016-09-17 19:55:10 +00:00
k:&:num, found?:bool <- next-ingredient
# we don't further check k for immutability, but assume it's mutable
]
$error: 0
:(scenario treat_optional_ingredients_as_mutable_2)
% Hide_errors = true;
def main [
local-scope
2016-09-17 19:55:10 +00:00
p:&:point <- new point:type
foo p
]
2016-09-17 19:55:10 +00:00
def foo p:&:point [
local-scope
load-ingredients
bar p
]
def bar [
local-scope
load-ingredients
2016-09-17 19:55:10 +00:00
p:&:point <- next-ingredient # optional ingredient; assumed to be mutable
]
+error: foo: cannot modify 'p' in instruction 'bar p' because it's an ingredient of recipe foo but not also a product
//: when checking for immutable ingredients, remember to take space into account
:(scenario check_space_of_reagents_in_immutability_checks)
def main [
2016-09-17 21:53:00 +00:00
a:space <- new-closure
2016-09-17 19:55:10 +00:00
b:&:num <- new number:type
2016-09-17 21:53:00 +00:00
run-closure b:&:num, a:space
]
def new-closure [
new-default-space
2016-09-17 19:55:10 +00:00
x:&:num <- new number:type
return default-space
]
2016-09-17 21:53:00 +00:00
def run-closure x:&:num, s:space [
local-scope
load-ingredients
2016-09-17 21:53:00 +00:00
0:space/names:new-closure <- copy s
# different space; always mutable
2016-09-17 19:55:10 +00:00
*x:&:num/space:1 <- copy 34
]
$error: 0
:(before "End Transforms")
Transform.push_back(check_immutable_ingredients); // idempotent
:(code)
2016-10-22 23:10:23 +00:00
void check_immutable_ingredients(const recipe_ordinal r) {
// to ensure an address reagent isn't modified, it suffices to show that
// a) we never write to its contents directly,
// b) we never call 'put' or 'put-index' on it, and
// c) any non-primitive recipe calls in the body aren't returning it as a product
const recipe& caller = get(Recipe, r);
trace(9991, "transform") << "--- check mutability of ingredients in recipe " << caller.name << end();
if (!caller.has_header) return; // skip check for old-style recipes calling next-ingredient directly
2016-10-20 05:10:35 +00:00
for (int i = 0; i < SIZE(caller.ingredients); ++i) {
const reagent& current_ingredient = caller.ingredients.at(i);
if (is_present_in_products(caller, current_ingredient.name)) continue; // not expected to be immutable
// End Immutable Ingredients Special-cases
set<reagent> immutable_vars;
immutable_vars.insert(current_ingredient);
2016-10-20 05:10:35 +00:00
for (int i = 0; i < SIZE(caller.steps); ++i) {
const instruction& inst = caller.steps.at(i);
check_immutable_ingredient_in_instruction(inst, immutable_vars, current_ingredient.name, caller);
2016-11-26 01:44:15 +00:00
if (inst.operation == INDEX && SIZE(inst.ingredients) > 1 && inst.ingredients.at(1).name == current_ingredient.name) continue;
update_aliases(inst, immutable_vars);
}
}
}
void update_aliases(const instruction& inst, set<reagent>& current_ingredient_and_aliases) {
set<int> current_ingredient_indices = ingredient_indices(inst, current_ingredient_and_aliases);
if (!contains_key(Recipe, inst.operation)) {
// primitive recipe
switch (inst.operation) {
case COPY:
2016-10-20 05:10:35 +00:00
for (set<int>::iterator p = current_ingredient_indices.begin(); p != current_ingredient_indices.end(); ++p)
current_ingredient_and_aliases.insert(inst.products.at(*p).name);
break;
case GET:
case INDEX:
case MAYBE_CONVERT:
// current_ingredient_indices can only have 0 or one value
2016-11-26 01:44:15 +00:00
if (!current_ingredient_indices.empty() && !inst.products.empty()) {
if (is_mu_address(inst.products.at(0)) || is_mu_container(inst.products.at(0)) || is_mu_exclusive_container(inst.products.at(0)))
current_ingredient_and_aliases.insert(inst.products.at(0));
}
break;
default: break;
}
}
else {
// defined recipe
set<int> contained_in_product_indices = scan_contained_in_product_indices(inst, current_ingredient_indices);
2016-10-20 05:10:35 +00:00
for (set<int>::iterator p = contained_in_product_indices.begin(); p != contained_in_product_indices.end(); ++p) {
if (*p < SIZE(inst.products))
current_ingredient_and_aliases.insert(inst.products.at(*p));
}
}
}
set<int> scan_contained_in_product_indices(const instruction& inst, set<int>& ingredient_indices) {
set<reagent> selected_ingredients;
const recipe& callee = get(Recipe, inst.operation);
2016-10-20 05:10:35 +00:00
for (set<int>::iterator p = ingredient_indices.begin(); p != ingredient_indices.end(); ++p) {
if (*p >= SIZE(callee.ingredients)) continue; // optional immutable ingredient
selected_ingredients.insert(callee.ingredients.at(*p));
}
set<int> result;
2016-10-20 05:10:35 +00:00
for (int i = 0; i < SIZE(callee.products); ++i) {
const reagent& current_product = callee.products.at(i);
const string_tree* contained_in_name = property(current_product, "contained-in");
if (contained_in_name && selected_ingredients.find(contained_in_name->value) != selected_ingredients.end())
result.insert(i);
}
return result;
}
:(scenarios transform)
:(scenario immutability_infects_contained_in_variables)
% Hide_errors = true;
container test-list [
2016-09-17 17:28:25 +00:00
value:num
2016-09-17 19:55:10 +00:00
next:&:test-list
]
def main [
local-scope
2016-09-17 19:55:10 +00:00
p:&:test-list <- new test-list:type
foo p
]
2016-09-17 19:55:10 +00:00
def foo p:&:test-list [ # p is immutable
local-scope
load-ingredients
2016-09-17 19:55:10 +00:00
p2:&:test-list <- test-next p # p2 is immutable
*p2 <- put *p2, value:offset, 34
]
2016-09-17 19:55:10 +00:00
def test-next x:&:test-list -> y:&:test-list/contained-in:x [
local-scope
load-ingredients
y <- get *x, next:offset
]
+error: foo: cannot modify 'p2' in instruction '*p2 <- put *p2, value:offset, 34' because that would modify p which is an ingredient of recipe foo but not also a product
:(code)
void check_immutable_ingredient_in_instruction(const instruction& inst, const set<reagent>& current_ingredient_and_aliases, const string& original_ingredient_name, const recipe& caller) {
// first check if the instruction is directly modifying something it shouldn't
2016-10-20 05:10:35 +00:00
for (int i = 0; i < SIZE(inst.products); ++i) {
if (has_property(inst.products.at(i), "lookup")
&& current_ingredient_and_aliases.find(inst.products.at(i)) != current_ingredient_and_aliases.end()) {
string current_product_name = inst.products.at(i).name;
if (current_product_name == original_ingredient_name)
raise << maybe(caller.name) << "cannot modify '" << current_product_name << "' in instruction '" << inst.original_string << "' because it's an ingredient of recipe " << caller.name << " but not also a product\n" << end();
else
raise << maybe(caller.name) << "cannot modify '" << current_product_name << "' in instruction '" << inst.original_string << "' because that would modify " << original_ingredient_name << " which is an ingredient of recipe " << caller.name << " but not also a product\n" << end();
return;
}
}
// check if there's any indirect modification going on
set<int> current_ingredient_indices = ingredient_indices(inst, current_ingredient_and_aliases);
if (current_ingredient_indices.empty()) return; // ingredient not found in call
2016-10-20 05:10:35 +00:00
for (set<int>::iterator p = current_ingredient_indices.begin(); p != current_ingredient_indices.end(); ++p) {
const int current_ingredient_index = *p;
reagent current_ingredient = inst.ingredients.at(current_ingredient_index);
canonize_type(current_ingredient);
const string& current_ingredient_name = current_ingredient.name;
if (!contains_key(Recipe, inst.operation)) {
// primitive recipe
// we got here only because we got an instruction with an implicit product, and the instruction didn't explicitly spell it out
// put x, y:offset, z
// instead of
// x <- put x, y:offset, z
if (inst.operation == PUT || inst.operation == PUT_INDEX) {
if (current_ingredient_index == 0) {
if (current_ingredient_name == original_ingredient_name)
raise << maybe(caller.name) << "cannot modify '" << current_ingredient_name << "' in instruction '" << inst.original_string << "' because it's an ingredient of recipe " << caller.name << " but not also a product\n" << end();
else
raise << maybe(caller.name) << "cannot modify '" << current_ingredient_name << "' in instruction '" << inst.original_string << "' because that would modify '" << original_ingredient_name << "' which is an ingredient of recipe " << caller.name << " but not also a product\n" << end();
}
}
}
else {
// defined recipe
if (is_modified_in_recipe(inst.operation, current_ingredient_index, caller)) {
if (current_ingredient_name == original_ingredient_name)
raise << maybe(caller.name) << "cannot modify '" << current_ingredient_name << "' in instruction '" << inst.original_string << "' because it's an ingredient of recipe " << caller.name << " but not also a product\n" << end();
else
raise << maybe(caller.name) << "cannot modify '" << current_ingredient_name << "' in instruction '" << inst.original_string << "' because that would modify '" << original_ingredient_name << "' which is an ingredient of recipe " << caller.name << " but not also a product\n" << end();
}
}
}
}
2016-10-22 23:10:23 +00:00
bool is_modified_in_recipe(const recipe_ordinal r, const int ingredient_index, const recipe& caller) {
const recipe& callee = get(Recipe, r);
if (!callee.has_header) {
raise << maybe(caller.name) << "can't check mutability of ingredients in recipe " << callee.name << " because it uses 'next-ingredient' directly, rather than a recipe header.\n" << end();
return true;
}
if (ingredient_index >= SIZE(callee.ingredients)) return false; // optional immutable ingredient
return is_present_in_products(callee, callee.ingredients.at(ingredient_index).name);
}
bool is_present_in_products(const recipe& callee, const string& ingredient_name) {
2016-10-20 05:10:35 +00:00
for (int i = 0; i < SIZE(callee.products); ++i) {
if (callee.products.at(i).name == ingredient_name)
return true;
}
return false;
}
set<int> ingredient_indices(const instruction& inst, const set<reagent>& ingredient_names) {
set<int> result;
2016-10-20 05:10:35 +00:00
for (int i = 0; i < SIZE(inst.ingredients); ++i) {
if (is_literal(inst.ingredients.at(i))) continue;
if (ingredient_names.find(inst.ingredients.at(i)) != ingredient_names.end())
result.insert(i);
}
return result;
}
//: Sometimes you want to pass in two addresses, one pointing inside the
//: other. For example, you want to delete a node from a linked list. You
//: can't pass both pointers back out, because if a caller tries to make both
//: identical then you can't tell which value will be written on the way out.
//:
2016-10-22 23:56:07 +00:00
//: Experimental solution: just tell Mu that one points inside the other.
//: This way we can return just one pointer as high up as necessary to capture
//: all modifications performed by a recipe.
//:
//: We'll see if we end up wanting to abuse /contained-in for other reasons.
:(scenarios transform)
:(scenario can_modify_contained_in_addresses)
container test-list [
2016-09-17 17:28:25 +00:00
value:num
2016-09-17 19:55:10 +00:00
next:&:test-list
]
def main [
local-scope
2016-09-17 19:55:10 +00:00
p:&:test-list <- new test-list:type
foo p
]
2016-09-17 19:55:10 +00:00
def foo p:&:test-list -> p:&:test-list [
local-scope
load-ingredients
2016-09-17 19:55:10 +00:00
p2:&:test-list <- test-next p
p <- test-remove p2, p
]
2016-09-17 19:55:10 +00:00
def test-next x:&:test-list -> y:&:test-list [
local-scope
load-ingredients
y <- get *x, next:offset
]
2016-09-17 19:55:10 +00:00
def test-remove x:&:test-list/contained-in:from, from:&:test-list -> from:&:test-list [
local-scope
load-ingredients
*x <- put *x, value:offset, 34 # can modify x
]
$error: 0
:(before "End Immutable Ingredients Special-cases")
if (has_property(current_ingredient, "contained-in")) {
const string_tree* tmp = property(current_ingredient, "contained-in");
3309 Rip out everything to fix one failing unit test (commit 3290; type abbreviations). This commit does several things at once that I couldn't come up with a clean way to unpack: A. It moves to a new representation for type trees without changing the actual definition of the `type_tree` struct. B. It adds unit tests for our type metadata precomputation, so that errors there show up early and in a simpler setting rather than dying when we try to load Mu code. C. It fixes a bug, guarding against infinite loops when precomputing metadata for recursive shape-shifting containers. To do this it uses a dumb way of comparing type_trees, comparing their string representations instead. That is likely incredibly inefficient. Perhaps due to C, this commit has made Mu incredibly slow. Running all tests for the core and the edit/ app now takes 6.5 minutes rather than 3.5 minutes. == more notes and details I've been struggling for the past week now to back out of a bad design decision, a premature optimization from the early days: storing atoms directly in the 'value' slot of a cons cell rather than creating a special 'atom' cons cell and storing it on the 'left' slot. In other words, if a cons cell looks like this: o / | \ left val right ..then the type_tree (a b c) used to look like this (before this commit): o | \ a o | \ b o | \ c null ..rather than like this 'classic' approach to s-expressions which never mixes val and right (which is what we now have): o / \ o o | / \ a o o | / \ b o null | c The old approach made several operations more complicated, most recently the act of replacing a (possibly atom/leaf) sub-tree with another. That was the final straw that got me to realize the contortions I was going through to save a few type_tree nodes (cons cells). Switching to the new approach was hard partly because I've been using the old approach for so long and type_tree manipulations had pervaded everything. Another issue I ran into was the realization that my layers were not cleanly separated. Key parts of early layers (precomputing type metadata) existed purely for far later ones (shape-shifting types). Layers I got repeatedly stuck at: 1. the transform for precomputing type sizes (layer 30) 2. type-checks on merge instructions (layer 31) 3. the transform for precomputing address offsets in types (layer 36) 4. replace operations in supporting shape-shifting recipes (layer 55) After much thrashing I finally noticed that it wasn't the entirety of these layers that was giving me trouble, but just the type metadata precomputation, which had bugs that weren't manifesting until 30 layers later. Or, worse, when loading .mu files before any tests had had a chance to run. A common failure mode was running into types at run time that I hadn't precomputed metadata for at transform time. Digging into these bugs got me to realize that what I had before wasn't really very good, but a half-assed heuristic approach that did a whole lot of extra work precomputing metadata for utterly meaningless types like `((address number) 3)` which just happened to be part of a larger type like `(array (address number) 3)`. So, I redid it all. I switched the representation of types (because the old representation made unit tests difficult to retrofit) and added unit tests to the metadata precomputation. I also made layer 30 only do the minimal metadata precomputation it needs for the concepts introduced until then. In the process, I also made the precomputation more correct than before, and added hooks in the right place so that I could augment the logic when I introduced shape-shifting containers. == lessons learned There's several levels of hygiene when it comes to layers: 1. Every layer introduces precisely what it needs and in the simplest way possible. If I was building an app until just that layer, nothing would seem over-engineered. 2. Some layers are fore-shadowing features in future layers. Sometimes this is ok. For example, layer 10 foreshadows containers and arrays and so on without actually supporting them. That is a net win because it lets me lay out the core of Mu's data structures out in one place. But if the fore-shadowing gets too complex things get nasty. Not least because it can be hard to write unit tests for features before you provide the plumbing to visualize and manipulate them. 3. A layer is introducing features that are tested only in later layers. 4. A layer is introducing features with tests that are invalidated in later layers. (This I knew from early on to be an obviously horrendous idea.) Summary: avoid Level 2 (foreshadowing layers) as much as possible. Tolerate it indefinitely for small things where the code stays simple over time, but become strict again when things start to get more complex. Level 3 is mostly a net lose, but sometimes it can be expedient (a real case of the usually grossly over-applied term "technical debt"), and it's better than the conventional baseline of no layers and no scenarios. Just clean it up as soon as possible. Definitely avoid layer 4 at any time. == minor lessons Avoid unit tests for trivial things, write scenarios in context as much as possible. But within those margins unit tests are fine. Just introduce them before any scenarios (commit 3297). Reorganizing layers can be easy. Just merge layers for starters! Punt on resplitting them in some new way until you've gotten them to work. This is the wisdom of Refactoring: small steps. What made it hard was not wanting to merge *everything* between layer 30 and 55. The eventual insight was realizing I just need to move those two full-strength transforms and nothing else.
2016-09-10 01:32:52 +00:00
if (!tmp->atom
|| (!is_present_in_ingredients(caller, tmp->value)
&& !is_present_in_products(caller, tmp->value))) {
3309 Rip out everything to fix one failing unit test (commit 3290; type abbreviations). This commit does several things at once that I couldn't come up with a clean way to unpack: A. It moves to a new representation for type trees without changing the actual definition of the `type_tree` struct. B. It adds unit tests for our type metadata precomputation, so that errors there show up early and in a simpler setting rather than dying when we try to load Mu code. C. It fixes a bug, guarding against infinite loops when precomputing metadata for recursive shape-shifting containers. To do this it uses a dumb way of comparing type_trees, comparing their string representations instead. That is likely incredibly inefficient. Perhaps due to C, this commit has made Mu incredibly slow. Running all tests for the core and the edit/ app now takes 6.5 minutes rather than 3.5 minutes. == more notes and details I've been struggling for the past week now to back out of a bad design decision, a premature optimization from the early days: storing atoms directly in the 'value' slot of a cons cell rather than creating a special 'atom' cons cell and storing it on the 'left' slot. In other words, if a cons cell looks like this: o / | \ left val right ..then the type_tree (a b c) used to look like this (before this commit): o | \ a o | \ b o | \ c null ..rather than like this 'classic' approach to s-expressions which never mixes val and right (which is what we now have): o / \ o o | / \ a o o | / \ b o null | c The old approach made several operations more complicated, most recently the act of replacing a (possibly atom/leaf) sub-tree with another. That was the final straw that got me to realize the contortions I was going through to save a few type_tree nodes (cons cells). Switching to the new approach was hard partly because I've been using the old approach for so long and type_tree manipulations had pervaded everything. Another issue I ran into was the realization that my layers were not cleanly separated. Key parts of early layers (precomputing type metadata) existed purely for far later ones (shape-shifting types). Layers I got repeatedly stuck at: 1. the transform for precomputing type sizes (layer 30) 2. type-checks on merge instructions (layer 31) 3. the transform for precomputing address offsets in types (layer 36) 4. replace operations in supporting shape-shifting recipes (layer 55) After much thrashing I finally noticed that it wasn't the entirety of these layers that was giving me trouble, but just the type metadata precomputation, which had bugs that weren't manifesting until 30 layers later. Or, worse, when loading .mu files before any tests had had a chance to run. A common failure mode was running into types at run time that I hadn't precomputed metadata for at transform time. Digging into these bugs got me to realize that what I had before wasn't really very good, but a half-assed heuristic approach that did a whole lot of extra work precomputing metadata for utterly meaningless types like `((address number) 3)` which just happened to be part of a larger type like `(array (address number) 3)`. So, I redid it all. I switched the representation of types (because the old representation made unit tests difficult to retrofit) and added unit tests to the metadata precomputation. I also made layer 30 only do the minimal metadata precomputation it needs for the concepts introduced until then. In the process, I also made the precomputation more correct than before, and added hooks in the right place so that I could augment the logic when I introduced shape-shifting containers. == lessons learned There's several levels of hygiene when it comes to layers: 1. Every layer introduces precisely what it needs and in the simplest way possible. If I was building an app until just that layer, nothing would seem over-engineered. 2. Some layers are fore-shadowing features in future layers. Sometimes this is ok. For example, layer 10 foreshadows containers and arrays and so on without actually supporting them. That is a net win because it lets me lay out the core of Mu's data structures out in one place. But if the fore-shadowing gets too complex things get nasty. Not least because it can be hard to write unit tests for features before you provide the plumbing to visualize and manipulate them. 3. A layer is introducing features that are tested only in later layers. 4. A layer is introducing features with tests that are invalidated in later layers. (This I knew from early on to be an obviously horrendous idea.) Summary: avoid Level 2 (foreshadowing layers) as much as possible. Tolerate it indefinitely for small things where the code stays simple over time, but become strict again when things start to get more complex. Level 3 is mostly a net lose, but sometimes it can be expedient (a real case of the usually grossly over-applied term "technical debt"), and it's better than the conventional baseline of no layers and no scenarios. Just clean it up as soon as possible. Definitely avoid layer 4 at any time. == minor lessons Avoid unit tests for trivial things, write scenarios in context as much as possible. But within those margins unit tests are fine. Just introduce them before any scenarios (commit 3297). Reorganizing layers can be easy. Just merge layers for starters! Punt on resplitting them in some new way until you've gotten them to work. This is the wisdom of Refactoring: small steps. What made it hard was not wanting to merge *everything* between layer 30 and 55. The eventual insight was realizing I just need to move those two full-strength transforms and nothing else.
2016-09-10 01:32:52 +00:00
raise << maybe(caller.name) << "/contained-in can only point to another ingredient or product, but got '" << to_string(property(current_ingredient, "contained-in")) << "'\n" << end();
}
continue;
}
:(scenario contained_in_check)
container test-list [
value:num
next:&:test-list
]
def test-remove x:&:test-list/contained-in:result, from:&:test-list -> result:&:test-list [
local-scope
load-ingredients
result <- copy 0
]
$error: 0