mu/040brace.cc
Kartik Agaram 4a943d4ed3 5001 - drop the :(scenario) DSL
I've been saying for a while[1][2][3] that adding extra abstractions makes
things harder for newcomers, and adding new notations doubly so. And then
I notice this DSL in my own backyard. Makes me feel like a hypocrite.

[1] https://news.ycombinator.com/item?id=13565743#13570092
[2] https://lobste.rs/s/to8wpr/configuration_files_are_canary_warning
[3] https://lobste.rs/s/mdmcdi/little_languages_by_jon_bentley_1986#c_3miuf2

The implementation of the DSL was also highly hacky:

a) It was happening in the tangle/ tool, but was utterly unrelated to tangling
layers.

b) There were several persnickety constraints on the different kinds of
lines and the specific order they were expected in. I kept finding bugs
where the translator would silently do the wrong thing. Or the error messages
sucked, and readers may be stuck looking at the generated code to figure
out what happened. Fixing error messages would require a lot more code,
which is one of my arguments against DSLs in the first place: they may
be easy to implement, but they're hard to design to go with the grain of
the underlying platform. They require lots of iteration. Is that effort
worth prioritizing in this project?

On the other hand, the DSL did make at least some readers' life easier,
the ones who weren't immediately put off by having to learn a strange syntax.
There were fewer quotes to parse, fewer backslash escapes.

Anyway, since there are also people who dislike having to put up with strange
syntaxes, we'll call that consideration a wash and tear this DSL out.

---

This commit was sheer drudgery. Hopefully it won't need to be redone with
a new DSL because I grow sick of backslashes.
2019-03-12 19:14:12 -07:00

567 lines
14 KiB
C++

//: Structured programming
//:
//: Our jump recipes are quite inconvenient to use, so Mu provides a
//: lightweight tool called 'transform_braces' to work in a slightly more
//: convenient format with nested braces:
//:
//: {
//: some instructions
//: {
//: more instructions
//: }
//: }
//:
//: Braces are just labels, they require no special parsing. The pseudo
//: instructions 'loop' and 'break' jump to just after the enclosing '{' and
//: '}' respectively.
//:
//: Conditional and unconditional 'loop' and 'break' should give us 80% of the
//: benefits of the control-flow primitives we're used to in other languages,
//: like 'if', 'while', 'for', etc.
void test_brace_conversion() {
transform(
"def main [\n"
" {\n"
" break\n"
" 1:num <- copy 0\n"
" }\n"
"]\n"
);
CHECK_TRACE_CONTENTS(
"transform: --- transform braces for recipe main\n"
"transform: jump 1:offset\n"
"transform: copy ...\n"
);
}
:(before "End Instruction Modifying Transforms")
Transform.push_back(transform_braces); // idempotent
:(code)
void transform_braces(const recipe_ordinal r) {
const bool OPEN = false, CLOSE = true;
// use signed integer for step index because we'll be doing arithmetic on it
list<pair<bool/*OPEN/CLOSE*/, /*step*/int> > braces;
trace(101, "transform") << "--- transform braces for recipe " << get(Recipe, r).name << end();
for (int index = 0; index < SIZE(get(Recipe, r).steps); ++index) {
const instruction& inst = get(Recipe, r).steps.at(index);
if (inst.label == "{") {
trace(103, "transform") << maybe(get(Recipe, r).name) << "push (open, " << index << ")" << end();
braces.push_back(pair<bool,int>(OPEN, index));
}
if (inst.label == "}") {
trace(103, "transform") << "push (close, " << index << ")" << end();
braces.push_back(pair<bool,int>(CLOSE, index));
}
}
stack</*step*/int> open_braces;
for (int index = 0; index < SIZE(get(Recipe, r).steps); ++index) {
instruction& inst = get(Recipe, r).steps.at(index);
if (inst.label == "{") {
open_braces.push(index);
continue;
}
if (inst.label == "}") {
if (open_braces.empty()) {
raise << maybe(get(Recipe, r).name) << "unbalanced '}'\n" << end();
return;
}
open_braces.pop();
continue;
}
if (inst.is_label) continue;
if (inst.name != "loop"
&& inst.name != "loop-if"
&& inst.name != "loop-unless"
&& inst.name != "break"
&& inst.name != "break-if"
&& inst.name != "break-unless") {
trace(102, "transform") << inst.name << " ..." << end();
continue;
}
// check for errors
if (inst.name.find("-if") != string::npos || inst.name.find("-unless") != string::npos) {
if (inst.ingredients.empty()) {
raise << maybe(get(Recipe, r).name) << "'" << inst.name << "' expects 1 or 2 ingredients, but got none\n" << end();
continue;
}
}
// update instruction operation
string old_name = inst.name; // save a copy
if (inst.name.find("-if") != string::npos) {
inst.name = "jump-if";
inst.operation = JUMP_IF;
}
else if (inst.name.find("-unless") != string::npos) {
inst.name = "jump-unless";
inst.operation = JUMP_UNLESS;
}
else {
inst.name = "jump";
inst.operation = JUMP;
}
// check for explicitly provided targets
if (inst.name.find("-if") != string::npos || inst.name.find("-unless") != string::npos) {
// conditional branches check arg 1
if (SIZE(inst.ingredients) > 1 && is_literal(inst.ingredients.at(1))) {
trace(102, "transform") << inst.name << ' ' << inst.ingredients.at(1).name << ":offset" << end();
continue;
}
}
else {
// unconditional branches check arg 0
if (!inst.ingredients.empty() && is_literal(inst.ingredients.at(0))) {
trace(102, "transform") << "jump " << inst.ingredients.at(0).name << ":offset" << end();
continue;
}
}
// if implicit, compute target
reagent target(new type_tree("offset"));
target.set_value(0);
if (open_braces.empty())
raise << maybe(get(Recipe, r).name) << "'" << old_name << "' needs a '{' before\n" << end();
else if (old_name.find("loop") != string::npos)
target.set_value(open_braces.top()-index);
else // break instruction
target.set_value(matching_brace(open_braces.top(), braces, r) - index - 1);
inst.ingredients.push_back(target);
// log computed target
if (inst.name == "jump")
trace(102, "transform") << "jump " << no_scientific(target.value) << ":offset" << end();
else
trace(102, "transform") << inst.name << ' ' << inst.ingredients.at(0).name << ", " << no_scientific(target.value) << ":offset" << end();
}
}
// returns a signed integer not just so that we can return -1 but also to
// enable future signed arithmetic
int matching_brace(int index, const list<pair<bool, int> >& braces, recipe_ordinal r) {
int stacksize = 0;
for (list<pair<bool, int> >::const_iterator p = braces.begin(); p != braces.end(); ++p) {
if (p->second < index) continue;
stacksize += (p->first ? 1 : -1);
if (stacksize == 0) return p->second;
}
raise << maybe(get(Recipe, r).name) << "unbalanced '{'\n" << end();
return SIZE(get(Recipe, r).steps); // exit current routine
}
void test_loop() {
transform(
"def main [\n"
" 1:num <- copy 0\n"
" 2:num <- copy 0\n"
" {\n"
" 3:num <- copy 0\n"
" loop\n"
" }\n"
"]\n"
);
CHECK_TRACE_CONTENTS(
"transform: --- transform braces for recipe main\n"
"transform: copy ...\n"
"transform: copy ...\n"
"transform: copy ...\n"
"transform: jump -2:offset\n"
);
}
void test_break_empty_block() {
transform(
"def main [\n"
" 1:num <- copy 0\n"
" {\n"
" break\n"
" }\n"
"]\n"
);
CHECK_TRACE_CONTENTS(
"transform: --- transform braces for recipe main\n"
"transform: copy ...\n"
"transform: jump 0:offset\n"
);
}
void test_break_cascading() {
transform(
"def main [\n"
" 1:num <- copy 0\n"
" {\n"
" break\n"
" }\n"
" {\n"
" break\n"
" }\n"
"]\n"
);
CHECK_TRACE_CONTENTS(
"transform: --- transform braces for recipe main\n"
"transform: copy ...\n"
"transform: jump 0:offset\n"
"transform: jump 0:offset\n"
);
}
void test_break_cascading_2() {
transform(
"def main [\n"
" 1:num <- copy 0\n"
" 2:num <- copy 0\n"
" {\n"
" break\n"
" 3:num <- copy 0\n"
" }\n"
" {\n"
" break\n"
" }\n"
"]\n"
);
CHECK_TRACE_CONTENTS(
"transform: --- transform braces for recipe main\n"
"transform: copy ...\n"
"transform: copy ...\n"
"transform: jump 1:offset\n"
"transform: copy ...\n"
"transform: jump 0:offset\n"
);
}
void test_break_if() {
transform(
"def main [\n"
" 1:num <- copy 0\n"
" 2:num <- copy 0\n"
" {\n"
" break-if 2:num\n"
" 3:num <- copy 0\n"
" }\n"
" {\n"
" break\n"
" }\n"
"]\n"
);
CHECK_TRACE_CONTENTS(
"transform: --- transform braces for recipe main\n"
"transform: copy ...\n"
"transform: copy ...\n"
"transform: jump-if 2, 1:offset\n"
"transform: copy ...\n"
"transform: jump 0:offset\n"
);
}
void test_break_nested() {
transform(
"def main [\n"
" 1:num <- copy 0\n"
" {\n"
" 2:num <- copy 0\n"
" break\n"
" {\n"
" 3:num <- copy 0\n"
" }\n"
" 4:num <- copy 0\n"
" }\n"
"]\n"
);
CHECK_TRACE_CONTENTS(
"transform: jump 4:offset\n"
);
}
void test_break_nested_degenerate() {
transform(
"def main [\n"
" 1:num <- copy 0\n"
" {\n"
" 2:num <- copy 0\n"
" break\n"
" {\n"
" }\n"
" 4:num <- copy 0\n"
" }\n"
"]\n"
);
CHECK_TRACE_CONTENTS(
"transform: jump 3:offset\n"
);
}
void test_break_nested_degenerate_2() {
transform(
"def main [\n"
" 1:num <- copy 0\n"
" {\n"
" 2:num <- copy 0\n"
" break\n"
" {\n"
" }\n"
" }\n"
"]\n"
);
CHECK_TRACE_CONTENTS(
"transform: jump 2:offset\n"
);
}
void test_break_label() {
Hide_errors = true;
transform(
"def main [\n"
" 1:num <- copy 0\n"
" {\n"
" break +foo:offset\n"
" }\n"
"]\n"
);
CHECK_TRACE_CONTENTS(
"transform: jump +foo:offset\n"
);
}
void test_break_unless() {
transform(
"def main [\n"
" 1:num <- copy 0\n"
" 2:num <- copy 0\n"
" {\n"
" break-unless 2:num\n"
" 3:num <- copy 0\n"
" }\n"
"]\n"
);
CHECK_TRACE_CONTENTS(
"transform: --- transform braces for recipe main\n"
"transform: copy ...\n"
"transform: copy ...\n"
"transform: jump-unless 2, 1:offset\n"
"transform: copy ...\n"
);
}
void test_loop_unless() {
transform(
"def main [\n"
" 1:num <- copy 0\n"
" 2:num <- copy 0\n"
" {\n"
" loop-unless 2:num\n"
" 3:num <- copy 0\n"
" }\n"
"]\n"
);
CHECK_TRACE_CONTENTS(
"transform: --- transform braces for recipe main\n"
"transform: copy ...\n"
"transform: copy ...\n"
"transform: jump-unless 2, -1:offset\n"
"transform: copy ...\n"
);
}
void test_loop_nested() {
transform(
"def main [\n"
" 1:num <- copy 0\n"
" {\n"
" 2:num <- copy 0\n"
" {\n"
" 3:num <- copy 0\n"
" }\n"
" loop-if 4:bool\n"
" 5:num <- copy 0\n"
" }\n"
"]\n"
);
CHECK_TRACE_CONTENTS(
"transform: --- transform braces for recipe main\n"
"transform: jump-if 4, -5:offset\n"
);
}
void test_loop_label() {
transform(
"def main [\n"
" 1:num <- copy 0\n"
" +foo\n"
" 2:num <- copy 0\n"
"]\n"
);
CHECK_TRACE_CONTENTS(
"transform: --- transform braces for recipe main\n"
"transform: copy ...\n"
"transform: copy ...\n"
);
}
//: test how things actually run
void test_brace_conversion_and_run() {
run(
"def test-factorial [\n"
" 1:num <- copy 5\n"
" 2:num <- copy 1\n"
" {\n"
" 3:bool <- equal 1:num, 1\n"
" break-if 3:bool\n"
" 2:num <- multiply 2:num, 1:num\n"
" 1:num <- subtract 1:num, 1\n"
" loop\n"
" }\n"
" 4:num <- copy 2:num\n" // trigger a read
"]\n"
);
CHECK_TRACE_CONTENTS(
"mem: location 2 is 120\n"
);
}
void test_break_outside_braces_fails() {
Hide_errors = true;
run(
"def main [\n"
" break\n"
"]\n"
);
CHECK_TRACE_CONTENTS(
"error: main: 'break' needs a '{' before\n"
);
}
void test_break_conditional_without_ingredient_fails() {
Hide_errors = true;
run(
"def main [\n"
" {\n"
" break-if\n"
" }\n"
"]\n"
);
CHECK_TRACE_CONTENTS(
"error: main: 'break-if' expects 1 or 2 ingredients, but got none\n"
);
}
//: Using break we can now implement conditional returns.
void test_return_if() {
run(
"def main [\n"
" 1:num <- test1\n"
"]\n"
"def test1 [\n"
" return-if 0, 34\n"
" return 35\n"
"]\n"
);
CHECK_TRACE_CONTENTS(
"mem: storing 35 in location 1\n"
);
}
void test_return_if_2() {
run(
"def main [\n"
" 1:num <- test1\n"
"]\n"
"def test1 [\n"
" return-if 1, 34\n"
" return 35\n"
"]\n"
);
CHECK_TRACE_CONTENTS(
"mem: storing 34 in location 1\n"
);
}
:(before "End Rewrite Instruction(curr, recipe result)")
// rewrite 'return-if a, b, c, ...' to
// ```
// {
// break-unless a
// return b, c, ...
// }
// ```
if (curr.name == "return-if" || curr.name == "reply-if") {
if (curr.products.empty()) {
emit_return_block(result, "break-unless", curr);
curr.clear();
}
else {
raise << "'" << curr.name << "' never yields any products\n" << end();
}
}
// rewrite 'return-unless a, b, c, ...' to
// ```
// {
// break-if a
// return b, c, ...
// }
// ```
if (curr.name == "return-unless" || curr.name == "reply-unless") {
if (curr.products.empty()) {
emit_return_block(result, "break-if", curr);
curr.clear();
}
else {
raise << "'" << curr.name << "' never yields any products\n" << end();
}
}
:(code)
void emit_return_block(recipe& out, const string& break_command, const instruction& inst) {
const vector<reagent>& ingredients = inst.ingredients;
reagent/*copy*/ condition = ingredients.at(0);
vector<reagent> return_ingredients;
copy(++ingredients.begin(), ingredients.end(), inserter(return_ingredients, return_ingredients.end()));
// {
instruction open_label; open_label.is_label=true; open_label.label = "{";
out.steps.push_back(open_label);
// <break command> <condition>
instruction break_inst;
break_inst.operation = get(Recipe_ordinal, break_command);
break_inst.name = break_command;
break_inst.ingredients.push_back(condition);
out.steps.push_back(break_inst);
// return <return ingredients>
instruction return_inst;
return_inst.operation = get(Recipe_ordinal, "return");
return_inst.name = "return";
return_inst.ingredients.swap(return_ingredients);
return_inst.original_string = inst.original_string;
out.steps.push_back(return_inst);
// }
instruction close_label; close_label.is_label=true; close_label.label = "}";
out.steps.push_back(close_label);
}
//: Make sure these pseudo recipes get consistent numbers in all tests, even
//: though they aren't implemented. Allows greater flexibility in ordering
//: transforms.
:(before "End Primitive Recipe Declarations")
BREAK,
BREAK_IF,
BREAK_UNLESS,
LOOP,
LOOP_IF,
LOOP_UNLESS,
:(before "End Primitive Recipe Numbers")
put(Recipe_ordinal, "break", BREAK);
put(Recipe_ordinal, "break-if", BREAK_IF);
put(Recipe_ordinal, "break-unless", BREAK_UNLESS);
put(Recipe_ordinal, "loop", LOOP);
put(Recipe_ordinal, "loop-if", LOOP_IF);
put(Recipe_ordinal, "loop-unless", LOOP_UNLESS);
:(before "End Primitive Recipe Checks")
case BREAK: break;
case BREAK_IF: break;
case BREAK_UNLESS: break;
case LOOP: break;
case LOOP_IF: break;
case LOOP_UNLESS: break;