1
0
Fork 0
mirror of https://github.com/Reuh/anselme.git synced 2025-10-28 00:59:31 +00:00
This commit is contained in:
Étienne Fildadut 2020-05-24 20:31:09 +02:00
parent 7a5a05ff34
commit b233d7fa1e
138 changed files with 4369 additions and 1611 deletions

153
parser/common.lua Normal file
View file

@ -0,0 +1,153 @@
local expression
local escapeCache = {}
local common
common = {
--- valid identifier pattern
identifier_pattern = "[^%%%/%*%+%-%(%)%!%&%|%=%$%?%>%<%:%{%}%[%]%,]+",
--- escape a string to be used as an exact match pattern
escape = function(str)
if not escapeCache[str] then
escapeCache[str] = str:gsub("[^%w]", "%%%1")
end
return escapeCache[str]
end,
--- trim a string
trim = function(str)
return str:match("^%s*(.-)%s*$")
end,
--- split a string separated by .
split = function(str)
local address = {}
for name in str:gmatch("[^%.]+") do
table.insert(address, name)
end
return address
end,
--- find a variable/function in a list, going up through the namespace hierarchy
find = function(list, namespace, name)
local ns = common.split(namespace)
for i=#ns, 1, -1 do
local fqm = ("%s.%s"):format(table.concat(ns, ".", 1, i), name)
if list[fqm] then
return list[fqm], fqm
end
end
if list[name] then
return list[name], name
end
return nil, ("can't find %q in namespace %s"):format(name, namespace)
end,
--- transform an identifier into a clean version
format_identifier = function(identifier, state)
local r = identifier:gsub("[^%.]+", function(str)
str = common.trim(str)
return state.aliases[str] or str
end)
return r
end,
--- flatten a nested list expression into a list of expressions
flatten_list = function(list, t)
t = t or {}
if list.type == "list" then
table.insert(t, list.left)
common.flatten_list(list.right, t)
else
table.insert(t, list)
end
return t
end,
-- * list of strings and expressions
-- * nil, err: in case of error
parse_text = function(text, state, namespace)
local l = {}
while text:match("[^%{]+") do
local t, e = text:match("^([^%{]*)(.-)$")
-- text
if t ~= "" then table.insert(l, t) end
-- expr
if e:match("^{") then
local exp, rem = expression(e:gsub("^{", ""), state, namespace)
if not exp then return nil, rem end
if not rem:match("^%s*}") then return nil, ("expected closing } at end of expression before %q"):format(rem) end
table.insert(l, exp)
text = rem:match("^%s*}(.*)$")
else
break
end
end
return l
end,
-- find compatible function variant
-- * variant: if success
-- * nil, err: if error
find_function_variant = function(fqm, state, arg, explicit_call)
local err = ("function %q variant not found"):format(fqm)
local func = state.functions[fqm] or {}
local args = arg and common.flatten_list(arg) or {}
for _, variant in ipairs(func) do
local ok = true
local return_type = variant.return_type
if variant.arity then
local min, max
if type(variant.arity) == "table" then
min, max = variant.arity[1], variant.arity[2]
else
min, max = variant.arity, variant.arity
end
if #args < min or #args > max then
if min == max then
err = ("function %q expected %s arguments but received %s"):format(fqm, min, #args)
else
err = ("function %q expected between %s and %s arguments but received %s"):format(fqm, min, max, #args)
end
ok = false
end
end
if ok and variant.check then
local s, e = variant.check(state, args)
if not s then
err = e or ("function %q variant failed to check arguments"):format(fqm)
ok = false
end
return_type = s == true and return_type or s
end
if ok and variant.types then
for j, t in pairs(variant.types) do
if args[j] and args[j].return_type and args[j].return_type ~= t then
err = ("function %q expected a %s as argument %s but received a %s"):format(fqm, t, j, args[j].return_type)
ok = false
end
end
end
if ok then
if variant.rewrite then
local r, e = variant.rewrite(fqm, state, arg, explicit_call)
if not r then
err = e
ok = false
end
if ok then
return r
end
else
return {
type = "function",
return_type = return_type,
name = fqm,
explicit_call = explicit_call,
variant = variant,
argument = arg
}
end
end
end
return nil, err
end
}
package.loaded[...] = common
expression = require((...):gsub("common$", "expression"))
return common

247
parser/expression.lua Normal file
View file

@ -0,0 +1,247 @@
local identifier_pattern, format_identifier, find, escape, find_function_variant, parse_text
--- binop priority
local binops_prio = {
[1] = { ";" },
[2] = { ":=", "+=", "-=", "//=", "/=", "*=", "%=", "^=" },
[3] = { "," },
[4] = { "|", "&" },
[5] = { "!=", "=", ">=", "<=", "<", ">" },
[6] = { "+", "-" },
[7] = { "*", "//", "/", "%" },
[8] = {}, -- unary operators
[9] = { "^", ":" },
[10] = { "." }
}
-- unop priority
local unops_prio = {
[1] = {},
[2] = {},
[3] = {},
[4] = {},
[5] = {},
[6] = {},
[7] = {},
[8] = { "-", "!" },
[9] = {}
}
--- parse an expression
-- return expr, remaining if success
-- returns nil, err if error
local function expression(s, state, namespace, currentPriority, operatingOn)
s = s:match("^%s*(.*)$")
currentPriority = currentPriority or 0
if not operatingOn then
-- number
if s:match("^%d+%.%d*") or s:match("^%d*%.%d+") or s:match("^%d+") then
local d, r = s:match("^(%d*%.%d*)(.*)$")
if not d then
d, r = s:match("^(%d+)(.*)$")
end
return expression(r, state, namespace, currentPriority, {
type = "number",
return_type = "number",
value = tonumber(d)
})
-- string
elseif s:match("^%\"[^\"]*%\"") then
local d, r = s:match("^%\"([^\"]*)%\"(.*)$")
while d:match("\\$") and not d:match("\\\\$") do
local nd, nr = r:match("([^\"]*)%\"(.*)$")
if not nd then return nil, ("unfinished string near %q"):format(r) end
d, r = d:sub(1, -2) .. "\"" .. nd, nr
end
local l, e = parse_text(tostring(d), state, namespace)
if not l then return l, e end
return expression(r, state, namespace, currentPriority, {
type = "string",
return_type = "string",
value = l
})
-- paranthesis
elseif s:match("^%b()") then
local content, r = s:match("^(%b())(.*)$")
content = content:gsub("^%(", ""):gsub("%)$", "")
local exp, r_paren = expression(content, state, namespace)
if not exp then return nil, "invalid expression inside parentheses: "..r_paren end
if r_paren:match("[^%s]") then return nil, ("unexpected %q at end of parenthesis expression"):format(r_paren) end
return expression(r, state, namespace, currentPriority, {
type = "parentheses",
return_type = exp.return_type,
expression = exp
})
-- list parenthesis
elseif s:match("^%b[]") then
local content, r = s:match("^(%b[])(.*)$")
content = content:gsub("^%[", ""):gsub("%]$", "")
local exp
if content:match("[^%s]") then
local r_paren
exp, r_paren = expression(content, state, namespace)
if not exp then return nil, "invalid expression inside list parentheses: "..r_paren end
if r_paren:match("[^%s]") then return nil, ("unexpected %q at end of list parenthesis expression"):format(r_paren) end
end
return expression(r, state, namespace, currentPriority, {
type = "list_parentheses",
return_type = "list",
expression = exp
})
-- identifier
elseif s:match("^"..identifier_pattern) then
local name, r = s:match("^("..identifier_pattern..")(.-)$")
name = format_identifier(name, state)
-- functions
local funcs, ffqm = find(state.functions, namespace, name)
if funcs then
local args, explicit_call
if r:match("^%b()") then
explicit_call = true
local content, rem = r:match("^(%b())(.*)$")
content = content:gsub("^%(", ""):gsub("%)$", "")
r = rem
-- get arguments
if content:match("[^%s]") then
local err
args, err = expression(content, state, namespace)
if not args then return args, err end
end
end
-- find compatible variant
local variant, err = find_function_variant(ffqm, state, args, explicit_call)
if not variant then return variant, err end
return expression(r, state, namespace, currentPriority, variant)
end
-- variables
local var, vfqm = find(state.variables, namespace, name)
if var then
return expression(r, state, namespace, currentPriority, {
type = "variable",
return_type = var.type ~= "undefined argument" and var.type or nil,
name = vfqm
})
end
-- suffix call: detect if prefix is valid variable, suffix call is handled in the binop section below
local sname, suffix = name:match("^(.*)(%."..identifier_pattern..")$")
if sname then
local svar, svfqm = find(state.variables, namespace, sname)
if svar then
return expression(suffix..r, state, namespace, currentPriority, {
type = "variable",
return_type = svar.type ~= "undefined argument" and svar.type or nil,
name = svfqm
})
end
end
return nil, ("unknown identifier %q"):format(name)
end
-- unops
for prio, oplist in ipairs(unops_prio) do
for _, op in ipairs(oplist) do
local escaped = escape(op)
if s:match("^"..escaped) then
local right, r = expression(s:match("^"..escaped.."(.*)$"), state, namespace, prio)
if not right then return nil, ("invalid expression after unop %q: %s"):format(op, r) end
-- find variant
local variant, err = find_function_variant(op, state, right, true)
if not variant then return variant, err end
return expression(r, state, namespace, currentPriority, variant)
end
end
end
return nil, ("no valid expression before %q"):format(s)
else
-- binop
for prio, oplist in ipairs(binops_prio) do
if prio >= currentPriority then
for _, op in ipairs(oplist) do
local escaped = escape(op)
if s:match("^"..escaped) then
local sright = s:match("^"..escaped.."(.*)$")
-- suffix call
if op == "." and sright:match("^"..identifier_pattern) then
local name, r = sright:match("^("..identifier_pattern..")(.-)$")
name = format_identifier(name, state)
local funcs, ffqm = find(state.functions, namespace, name)
if funcs then
local args, explicit_call
if r:match("^%b()") then
explicit_call = true
local content, rem = r:match("^(%b())(.*)$")
content = content:gsub("^%(", ""):gsub("%)$", "")
r = rem
-- get arguments
if content:match("[^%s]") then
local err
args, err = expression(content, state, namespace)
if not args then return args, err end
end
end
-- add first argument
if not args then
args = operatingOn
else
args = {
type = "list",
return_type = "list",
left = operatingOn,
right = args
}
end
-- find compatible variant
local variant, err = find_function_variant(ffqm, state, args, explicit_call)
if not variant then return variant, err end
return expression(r, state, namespace, currentPriority, variant)
end
else
local right, r = expression(sright, state, namespace, prio)
if not right then return nil, ("invalid expression after binop %q: %s"):format(op, r) end
-- list constructor
if op == "," then
return expression(r, state, namespace, currentPriority, {
type = "list",
return_type = "list",
left = operatingOn,
right = right
})
-- normal binop
else
-- find variant
local args = {
type = "list",
return_type = "list",
left = operatingOn,
-- wrap in parentheses to avoid appending to argument list if right is a list
right = { type = "parentheses", return_type = right.return_type, expression = right }
}
local variant, err = find_function_variant(op, state, args, true)
if not variant then return variant, err end
return expression(r, state, namespace, currentPriority, variant)
end
end
end
end
end
end
-- index
if s:match("^%b()") then
local content, r = s:match("^(%b())(.*)$")
-- get arguments (parentheses are kept)
local right, r_paren = expression(content, state, namespace)
if not right then return right, r_paren end
if r_paren:match("[^%s]") then return nil, ("unexpected %q at end of index expression"):format(r_paren) end
local args = { type = "list", left = operatingOn, right = right }
local variant, err = find_function_variant("(", state, args, true)
if not variant then return variant, err end
return expression(r, state, namespace, currentPriority, variant)
end
-- nothing to operate
return operatingOn, s
end
end
package.loaded[...] = expression
local common = require((...):gsub("expression$", "common"))
identifier_pattern, format_identifier, find, escape, find_function_variant, parse_text = common.identifier_pattern, common.format_identifier, common.find, common.escape, common.find_function_variant, common.parse_text
return expression

70
parser/postparser.lua Normal file
View file

@ -0,0 +1,70 @@
local expression
local parse_text
-- * true: if success
-- * nil, error: in case of error
local function parse(state)
for _, l in ipairs(state.queued_lines) do
local line, namespace = l.line, l.namespace
-- decorators
if line.condition then
if line.condition:match("[^%s]") then
local exp, rem = expression(line.condition, state, namespace)
if not exp then return nil, ("%s; at line %s"):format(rem, line.line) end
if rem:match("[^%s]") then return nil, ("expected end of expression before %q in condition decorator; at line %s"):format(rem, line.line) end
line.condition = exp
else
line.condition = nil
end
end
if line.tag then
if line.tag:match("[^%s]") then
local exp, rem = expression(line.tag, state, namespace)
if not exp then return nil, ("%s; at line %s"):format(rem, line.line) end
if rem:match("[^%s]") then return nil, ("expected end of expression before %q in condition decorator; at line %s"):format(rem, line.line) end
line.tag = exp
else
line.tag = nil
end
end
-- expressions
if line.expression then
if line.expression:match("[^%s]") then
local exp, rem = expression(line.expression, state, namespace)
if not exp then return nil, ("%s; at line %s"):format(rem, line.line) end
if rem:match("[^%s]") then return nil, ("expected end of expression before %q; at line %s"):format(rem, line.line) end
line.expression = exp
else
line.expression = nil
end
-- function return type information
if line.type == "return" then
local variant = line.parent_function.variant
local return_type = line.expression.return_type
if return_type then
if not variant.return_type then
variant.return_type = return_type
elseif variant.return_type ~= return_type then
return nil, ("trying to return a %s in a function that returns a %s; at line %s"):format(return_type, variant.return_type, line.line)
end
end
end
end
-- text
if line.text then
local txt, err = parse_text(line.text, state, namespace)
if err then return nil, ("%s; at line %s"):format(err, line.line) end
line.text = txt
end
end
state.queued_lines = {}
return true
end
package.loaded[...] = parse
expression = require((...):gsub("postparser$", "expression"))
local common = require((...):gsub("postparser$", "common"))
parse_text = common.parse_text
--- postparse shit: parse expressions and do variable existence and type checking
return parse

332
parser/preparser.lua Normal file
View file

@ -0,0 +1,332 @@
local expression
local format_identifier, identifier_pattern
local eval
-- * ast: if success
-- * nil, error: in case of error
local function parse_line(line, state, namespace)
local l = line.content
local r = {
line = line.line
}
-- comment
if l:match("^%(") then
r.type = "comment"
r.remove_from_block_ast = true
return r
end
-- decorators
while l:match("^..+[~#]") or l:match("^..+§") do
-- condition
if l:match("^..+%~.-$") then
local expr
l, expr = l:match("^(.-)%s*%~(.-)$")
r.condition = expr
-- paragraph
elseif l:match("^..+§.-$") then
local name
l, name = l:match("^(.-)%s*§(.-)$")
local fqm = ("%s%s"):format(namespace, format_identifier(name, state))
namespace = fqm.."."
r.paragraph = true
r.parent_function = true
r.namespace = fqm.."."
r.name = fqm
if not state.functions[fqm] then
state.functions[fqm] = {
{
arity = 0,
value = r
}
}
if not state.variables[fqm..".👁️"] then
state.variables[fqm..".👁️"] = {
type = "number",
value = 0
}
end
else
table.insert(state.functions[fqm], {
arity = 0,
value = r
})
end
-- tag
elseif l:match("^..+%#.-$") then
local expr
l, expr = l:match("^(.-)%s*%#(.-)$")
r.tag = expr
end
end
-- else-condition & condition
if l:match("^~~?") then
r.type = l:match("^~~") and "else-condition" or "condition"
r.child = true
local expr = l:match("^~~?(.*)$")
if expr:match("[^%s]") then
r.expression = expr
else
r.expression = "1"
end
-- choice
elseif l:match("^>") then
r.type = "choice"
r.push_event = "choice"
r.child = true
r.text = l:match("^>%s*(.-)$")
-- function & paragraph
elseif l:match("^%$") or l:match("") then -- § is a 2-bytes caracter, DO NOT USE LUA PATTERN OPERATORS as they operate on single bytes
r.type = l:match("^%$") and "function" or "paragraph"
r.child = true
local fqm = ("%s%s"):format(namespace, format_identifier(l:match("^%$(.*)$") or l:match("^§(.*)$"), state))
-- get params
r.params = {}
if r.type == "function" and fqm:match("%b()$") then
local content
fqm, content = fqm:match("^(.-)(%b())$")
content = content:gsub("^%(", ""):gsub("%)$", "")
for param in content:gmatch("[^%,]+") do
table.insert(r.params, format_identifier(("%s.%s"):format(fqm, param), state))
end
end
local arity, vararg = #r.params, nil
if arity > 0 and r.params[arity]:match("%.%.%.$") then -- varargs
r.params[arity] = r.params[arity]:match("^(.*)%.%.%.$")
vararg = arity
arity = { arity-1, math.huge }
end
-- store parent function and run paragraph when line is read
if r.type == "paragraph" then
r.paragraph = true
r.parent_function = true
end
-- don't keep function node in block AST
if r.type == "function" then
r.remove_from_block_ast = true
if not state.variables[fqm..".🏁"] then
state.variables[fqm..".🏁"] = {
type = "string",
value = ""
}
end
end
-- define function and variables
r.namespace = fqm.."."
r.name = fqm
if state.variables[fqm] then return nil, ("trying to define %s %s, but a variable with the same name exists; at line %s"):format(r.type, fqm, line.line) end
r.variant = {
arity = arity,
types = {},
vararg = vararg,
value = r
}
if not state.functions[fqm] then
state.functions[fqm] = { r.variant }
if not state.variables[fqm..".👁️"] then
state.variables[fqm..".👁️"] = {
type = "number",
value = 0
}
end
else
-- check for arity conflict
for _, variant in ipairs(state.functions[fqm]) do
local vmin, vmax = 0, math.huge
if type(variant.arity) == "table" then
vmin, vmax = variant.arity[1], variant.arity[2]
elseif variant.arity then
vmin, vmax = variant.arity, variant.arity
end
local min, max = 0, math.huge
if type(r.variant.arity) == "table" then
min, max = r.variant.arity[1], r.variant.arity[2]
elseif r.variant.arity then
min, max = variant.arity, r.variant.arity
end
if min == vmin and max == vmax then
return nil, ("trying to define %s %s with arity [%s;%s], but another function with the arity exist; at line %s"):format(r.type, fqm, min, max, line.line)
end
end
-- add
table.insert(state.functions[fqm], r.variant)
end
-- set type check information
for i, param in ipairs(r.params) do
if not state.variables[param] then
state.variables[param] = {
type = "undefined argument",
value = { r.variant, i }
}
elseif state.variables[param].type ~= "undefined argument" then
r.variant.types[i] = state.variables[param].type
end
end
-- definition
elseif l:match("^:") then
r.type = "definition"
r.remove_from_block_ast = true
local exp, rem = expression(l:match("^:(.*)$"), state, namespace) -- expression parsing is done directly to get type information
if not exp then return nil, ("%s; at line %s"):format(rem, line.line) end
local fqm = ("%s%s"):format(namespace, format_identifier(rem, state))
if state.functions[fqm] then return nil, ("trying to define variable %s, but a function with the same name exists; at line %s"):format(fqm, line.line) end
if not state.variables[fqm] or state.variables[fqm].type == "undefined argument" then
local v, e = eval(state, exp)
if not v then return v, e end
-- update function typecheck information
if state.variables[fqm] and state.variables[fqm].type == "undefined argument" then
local und = state.variables[fqm].value
und[1].types[und[2]] = v.type
end
state.variables[fqm] = v
elseif state.variables[fqm].type ~= exp.type then
return nil, ("trying to define variable %s of type %s but a it is already defined with type %s; at line %s"):format(fqm, exp.type, state.variables[fqm].type, line.line)
end
-- tag
elseif l:match("^%#") then
r.type = "tag"
r.child = true
r.expression = l:match("^%#(.*)$")
-- return
elseif l:match("^%@") then
r.type = "return"
r.parent_function = true
r.expression = l:match("^%@(.*)$")
-- text
elseif l:match("[^%s]") then
r.type = "text"
r.push_event = "text"
r.text = l
-- flush events
else
r.type = "flush_events"
end
if not r.type then return nil, ("unknown line %s type"):format(line.line) end
return r
end
-- * block: in case of success
-- * nil, err: in case of error
local function parse_block(indented, state, namespace, parent_function, last_event)
local block = { type = "block" }
local lastLine -- last line AST
for i, l in ipairs(indented) do
-- parsable line
if l.content then
local ast, err = parse_line(l, state, namespace)
if err then return nil, err end
lastLine = ast
-- store parent function
if ast.parent_function then ast.parent_function = parent_function end
-- add to block AST
if not ast.remove_from_block_ast then
ast.parent_block = block
-- insert flush on event type change
if ast.type == "flush" then last_event = nil end
if ast.push_event then
if last_event and ast.push_event ~= last_event then
table.insert(block, { line = l.line, type = "flush_events" })
end
last_event = ast.push_event
end
-- add ast node
ast.parent_position = #block+1
if ast.replace_with then
if indented[i+1].content then
table.insert(indented, i+1, { content = ast.replace_with, line = l.line })
else
table.insert(indented, i+2, { content = ast.replace_with, line = l.line }) -- if line has children
end
else
table.insert(block, ast)
end
end
-- add child
if ast.child then ast.child = { type = "block", parent_line = ast } end
-- queue in expression evalution
table.insert(state.queued_lines, { namespace = ast.namespace or namespace, line = ast })
-- indented (ignore block comments)
elseif lastLine.type ~= "comment" then
if not lastLine.child then
return nil, ("line %s (%s) can't have children"):format(lastLine.line, lastLine.type)
else
local r, e = parse_block(l, state, lastLine.namespace or namespace, lastLine.type == "function" and lastLine or parent_function, last_event)
if not r then return r, e end
r.parent_line = lastLine
lastLine.child = r
end
end
end
return block
end
--- returns the nested list of lines {content="", line=1}, grouped by indentation
-- multiple empty lines are merged
-- * list, last line
local function parse_indent(lines, i, indentLevel, insert_empty_line)
i = i or 1
indentLevel = indentLevel or 0
local indented = {}
while i <= #lines do
if lines[i]:match("[^%s]") then
local indent, line = lines[i]:match("^(%s*)(.*)$")
if #indent == indentLevel then
if insert_empty_line then
table.insert(indented, { content = "", line = insert_empty_line })
insert_empty_line = false
end
table.insert(indented, { content = line, line = i })
elseif #indent > indentLevel then
local t
t, i = parse_indent(lines, i, #indent, insert_empty_line)
table.insert(indented, t)
else
return indented, i-1
end
elseif not insert_empty_line then
insert_empty_line = i
end
i = i + 1
end
return indented, i-1
end
--- return the list of raw lines of s
local function parse_lines(s)
local lines = {}
for l in (s.."\n"):gmatch("(.-)\n") do
table.insert(lines, l)
end
return lines
end
--- preparse shit: create AST structure, define variables and functions, but don't parse expression or perform any type checking
-- (wait for other files to be parsed before doing this with postparse)
-- * state: in case of success
-- * nil, err: in case of error
local function parse(state, s, name)
-- parse lines
local lines = parse_lines(s)
local indented = parse_indent(lines)
-- wrap in named function if neccessary
if name ~= "" then
if not name:match("^"..identifier_pattern.."$") then
return nil, ("invalid function name %q"):format(name)
end
indented = {
{ content = "$ "..name, line = 0 },
indented
}
end
-- parse
local root, err = parse_block(indented, state, "")
if not root then return nil, err end
return state
end
package.loaded[...] = parse
expression = require((...):gsub("preparser$", "expression"))
local common = require((...):gsub("preparser$", "common"))
format_identifier, identifier_pattern = common.format_identifier, common.identifier_pattern
eval = require((...):gsub("parser%.preparser$", "interpreter.expression"))
return parse