mu/031merge.cc

228 lines
8.0 KiB
C++
Raw Normal View History

2016-05-03 16:19:58 +00:00
//: Construct types out of their constituent fields.
:(scenario merge)
container foo [
x:num
y:num
2016-05-03 16:19:58 +00:00
]
def main [
1:foo <- merge 3, 4
]
+mem: storing 3 in location 1
+mem: storing 4 in location 2
:(before "End Primitive Recipe Declarations")
MERGE,
:(before "End Primitive Recipe Numbers")
put(Recipe_ordinal, "merge", MERGE);
:(before "End Primitive Recipe Checks")
case MERGE: {
// type-checking in a separate transform below
break;
}
:(before "End Primitive Recipe Implementations")
case MERGE: {
products.resize(1);
2016-10-20 05:10:35 +00:00
for (int i = 0; i < SIZE(ingredients); ++i)
for (int j = 0; j < SIZE(ingredients.at(i)); ++j)
2016-05-03 16:19:58 +00:00
products.at(0).push_back(ingredients.at(i).at(j));
break;
}
//: type-check 'merge' to avoid interpreting numbers as addresses
:(scenario merge_check)
def main [
1:point <- merge 3, 4
]
$error: 0
:(scenario merge_check_missing_element)
% Hide_errors = true;
def main [
1:point <- merge 3
]
+error: main: too few ingredients in '1:point <- merge 3'
:(scenario merge_check_extra_element)
% Hide_errors = true;
def main [
1:point <- merge 3, 4, 5
]
+error: main: too many ingredients in '1:point <- merge 3, 4, 5'
//: We want to avoid causing memory corruption, but other than that we want to
//: be flexible in how we construct containers of containers. It should be
//: equally easy to define a container out of primitives or intermediate
//: container fields.
:(scenario merge_check_recursive_containers)
def main [
1:point <- merge 3, 4
1:point-number <- merge 1:point, 5
]
$error: 0
:(scenario merge_check_recursive_containers_2)
% Hide_errors = true;
def main [
1:point <- merge 3, 4
2:point-number <- merge 1:point
]
+error: main: too few ingredients in '2:point-number <- merge 1:point'
:(scenario merge_check_recursive_containers_3)
def main [
1:point-number <- merge 3, 4, 5
]
$error: 0
:(scenario merge_check_recursive_containers_4)
% Hide_errors = true;
def main [
1:point-number <- merge 3, 4
]
+error: main: too few ingredients in '1:point-number <- merge 3, 4'
:(scenario merge_check_reflexive)
% Hide_errors = true;
def main [
1:point <- merge 3, 4
2:point <- merge 1:point
]
$error: 0
//: Since a container can be merged in several ways, we need to be able to
//: backtrack through different possibilities. Later we'll allow creating
//: exclusive containers which contain just one of rather than all of their
//: elements. That will also require backtracking capabilities. Here's the
//: state we need to maintain for backtracking:
:(before "End Types")
struct merge_check_point {
reagent container;
int container_element_index;
merge_check_point(const reagent& c, int i) :container(c), container_element_index(i) {}
};
struct merge_check_state {
stack<merge_check_point> data;
};
:(before "End Checks")
Transform.push_back(check_merge_calls); // idempotent
2016-05-03 16:19:58 +00:00
:(code)
void check_merge_calls(const recipe_ordinal r) {
const recipe& caller = get(Recipe, r);
trace(9991, "transform") << "--- type-check merge instructions in recipe " << caller.name << end();
2016-10-20 05:10:35 +00:00
for (int i = 0; i < SIZE(caller.steps); ++i) {
2016-05-03 16:19:58 +00:00
const instruction& inst = caller.steps.at(i);
if (inst.name != "merge") continue;
if (SIZE(inst.products) != 1) {
2017-05-26 23:43:18 +00:00
raise << maybe(caller.name) << "'merge' should yield a single product in '" << to_original_string(inst) << "'\n" << end();
2016-05-03 16:19:58 +00:00
continue;
}
2016-05-06 07:46:39 +00:00
reagent/*copy*/ product = inst.products.at(0);
2016-05-03 16:19:58 +00:00
// Update product While Type-checking Merge
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
const type_tree* product_base_type = product.type->atom ? product.type : product.type->left;
assert(product_base_type->atom);
if (product_base_type->value == 0 || !contains_key(Type, product_base_type->value)) {
2017-05-26 23:43:18 +00:00
raise << maybe(caller.name) << "'merge' should yield a container in '" << to_original_string(inst) << "'\n" << end();
2016-05-03 16:19:58 +00:00
continue;
}
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
const type_info& info = get(Type, product_base_type->value);
2016-05-03 16:19:58 +00:00
if (info.kind != CONTAINER && info.kind != EXCLUSIVE_CONTAINER) {
2017-05-26 23:43:18 +00:00
raise << maybe(caller.name) << "'merge' should yield a container in '" << to_original_string(inst) << "'\n" << end();
2016-05-03 16:19:58 +00:00
continue;
}
check_merge_call(inst.ingredients, product, caller, inst);
}
}
void check_merge_call(const vector<reagent>& ingredients, const reagent& product, const recipe& caller, const instruction& inst) {
int ingredient_index = 0;
merge_check_state state;
state.data.push(merge_check_point(product, 0));
while (true) {
assert(!state.data.empty());
trace("transform") << ingredient_index << " vs " << SIZE(ingredients) << end();
2016-05-03 16:19:58 +00:00
if (ingredient_index >= SIZE(ingredients)) {
2017-05-26 23:43:18 +00:00
raise << maybe(caller.name) << "too few ingredients in '" << to_original_string(inst) << "'\n" << end();
2016-05-03 16:19:58 +00:00
return;
}
reagent& container = state.data.top().container;
2016-06-18 00:22:15 +00:00
if (!container.type) return; // error handled elsewhere
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
const type_tree* top_root_type = container.type->atom ? container.type : container.type->left;
assert(top_root_type->atom);
type_info& container_info = get(Type, top_root_type->value);
2016-05-03 16:19:58 +00:00
switch (container_info.kind) {
case CONTAINER: {
// degenerate case: merge with the same type always succeeds
if (state.data.top().container_element_index == 0 && types_coercible(container, inst.ingredients.at(ingredient_index)))
return;
2016-05-06 07:46:39 +00:00
const reagent& expected_ingredient = element_type(container.type, state.data.top().container_element_index);
trace("transform") << "checking container " << to_string(container) << " || " << to_string(expected_ingredient) << " vs ingredient " << ingredient_index << end();
2016-05-03 16:19:58 +00:00
// if the current element is the ingredient we expect, move on to the next element/ingredient
if (types_coercible(expected_ingredient, ingredients.at(ingredient_index))) {
++ingredient_index;
++state.data.top().container_element_index;
while (state.data.top().container_element_index >= SIZE(get(Type, get_base_type(state.data.top().container.type)->value).elements)) {
2016-05-03 16:19:58 +00:00
state.data.pop();
if (state.data.empty()) {
if (ingredient_index < SIZE(ingredients))
2017-05-26 23:43:18 +00:00
raise << maybe(caller.name) << "too many ingredients in '" << to_original_string(inst) << "'\n" << end();
2016-05-03 16:19:58 +00:00
return;
}
++state.data.top().container_element_index;
}
}
// if not, maybe it's a field of the current element
else {
// no change to ingredient_index
state.data.push(merge_check_point(expected_ingredient, 0));
}
break;
}
// End check_merge_call Special-cases
2016-05-03 16:19:58 +00:00
default: {
if (!types_coercible(container, ingredients.at(ingredient_index))) {
2017-05-26 23:43:18 +00:00
raise << maybe(caller.name) << "incorrect type of ingredient " << ingredient_index << " in '" << to_original_string(inst) << "'\n" << end();
raise << " (expected '" << debug_string(container) << "')\n" << end();
raise << " (got '" << debug_string(ingredients.at(ingredient_index)) << "')\n" << end();
2016-05-03 16:19:58 +00:00
return;
}
++ingredient_index;
// ++state.data.top().container_element_index; // unnecessary, but wouldn't do any harm
do {
state.data.pop();
if (state.data.empty()) {
if (ingredient_index < SIZE(ingredients))
2017-05-26 23:43:18 +00:00
raise << maybe(caller.name) << "too many ingredients in '" << to_original_string(inst) << "'\n" << end();
2016-05-03 16:19:58 +00:00
return;
}
++state.data.top().container_element_index;
} while (state.data.top().container_element_index >= SIZE(get(Type, get_base_type(state.data.top().container.type)->value).elements));
2016-05-03 16:19:58 +00:00
}
}
}
// never gets here
assert(false);
}
//: replaced in a later layer
//: todo: find some clean way to take this call completely out of this layer
const type_tree* get_base_type(const type_tree* t) {
return t;
}
2016-05-03 16:19:58 +00:00
:(scenario merge_check_product)
% Hide_errors = true;
def main [
1:num <- merge 3
2016-05-03 16:19:58 +00:00
]
+error: main: 'merge' should yield a container in '1:num <- merge 3'
2016-05-03 16:19:58 +00:00
:(before "End Includes")
#include <stack>
using std::stack;