Move protocol code generators to liberty

This part of the project is now more or less stable.
This commit is contained in:
Přemysl Eric Janouch 2022-09-30 03:23:06 +02:00
parent 86278c154c
commit 46d68eacce
Signed by: p
GPG Key ID: A0420B94F92B9493
8 changed files with 14 additions and 1407 deletions

View File

@ -162,15 +162,15 @@ add_custom_target (replies DEPENDS ${PROJECT_BINARY_DIR}/xD-replies.c)
add_custom_command (OUTPUT xC-proto.c
COMMAND env LC_ALL=C awk
-f ${PROJECT_SOURCE_DIR}/xC-gen-proto.awk
-f ${PROJECT_SOURCE_DIR}/xC-gen-proto-c.awk
-f ${PROJECT_SOURCE_DIR}/liberty/tools/lxdrgen.awk
-f ${PROJECT_SOURCE_DIR}/liberty/tools/lxdrgen-c.awk
-v PrefixCamel=Relay
${PROJECT_SOURCE_DIR}/xC-proto > xC-proto.c
${PROJECT_SOURCE_DIR}/xC.lxdr > xC-proto.c
DEPENDS
${PROJECT_SOURCE_DIR}/xC-gen-proto.awk
${PROJECT_SOURCE_DIR}/xC-gen-proto-c.awk
${PROJECT_SOURCE_DIR}/xC-proto
COMMENT "Generating xC relay protocol code")
${PROJECT_SOURCE_DIR}/liberty/tools/lxdrgen.awk
${PROJECT_SOURCE_DIR}/liberty/tools/lxdrgen-c.awk
${PROJECT_SOURCE_DIR}/xC.lxdr
COMMENT "Generating xC relay protocol code" VERBATIM)
add_custom_target (xC-proto DEPENDS ${PROJECT_BINARY_DIR}/xC-proto.c)
# Build

@ -1 +1 @@
Subproject commit af2756ee01fa6b1921c6bcb581817e64c30beb48
Subproject commit 035bfe5e81b80ef9df03414c7c567093ce26629a

View File

@ -1,324 +0,0 @@
# xC-gen-proto-c.awk: C backend for xC-gen-proto.awk.
#
# Copyright (c) 2022, Přemysl Eric Janouch <p@janouch.name>
# SPDX-License-Identifier: 0BSD
#
# Neither *_new() nor *_destroy() functions are provided, because they'd only
# be useful for top-levels, and are merely extra malloc()/free() calls.
# Users are expected to reuse buffers.
#
# Similarly, no constructors are produced--those are easy to write manually.
#
# All arrays are deserialized zero-terminated, so u8<> and i8<> can be directly
# used as C strings.
#
# All types must be able to dispose partially zero values going from the back,
# i.e., in the reverse order of deserialization.
function define_internal(name, ctype) {
Types[name] = "internal"
CodegenCType[name] = ctype
}
function define_int(shortname, ctype) {
define_internal(shortname, ctype)
CodegenSerialize[shortname] = \
"\tstr_pack_" shortname "(w, %s);\n"
CodegenDeserialize[shortname] = \
"\tif (!msg_unpacker_" shortname "(r, &%s))\n" \
"\t\treturn false;\n"
}
function define_sint(size) { define_int("i" size, "int" size "_t") }
function define_uint(size) { define_int("u" size, "uint" size "_t") }
function codegen_begin() {
define_sint("8")
define_sint("16")
define_sint("32")
define_sint("64")
define_uint("8")
define_uint("16")
define_uint("32")
define_uint("64")
define_internal("string", "struct str")
CodegenDispose["string"] = "\tstr_free(&%s);\n"
CodegenSerialize["string"] = \
"\tif (!proto_string_serialize(&%s, w))\n" \
"\t\treturn false;\n"
CodegenDeserialize["string"] = \
"\tif (!proto_string_deserialize(&%s, r))\n" \
"\t\treturn false;\n"
define_internal("bool", "bool")
CodegenSerialize["bool"] = \
"\tstr_pack_u8(w, !!%s);\n"
CodegenDeserialize["bool"] = \
"\t{\n" \
"\t\tuint8_t v = 0;\n" \
"\t\tif (!msg_unpacker_u8(r, &v))\n" \
"\t\t\treturn false;\n" \
"\t\t%s = !!v;\n" \
"\t}\n"
print "// Code generated from " FILENAME ". DO NOT EDIT."
print "// This file directly depends on liberty.c, but doesn't include it."
print ""
print "static bool"
print "proto_string_serialize(const struct str *s, struct str *w) {"
print "\tif (s->len > UINT32_MAX)"
print "\t\treturn false;"
print "\tstr_pack_u32(w, s->len);"
print "\tstr_append_str(w, s);"
print "\treturn true;"
print "}"
print ""
print "static bool"
print "proto_string_deserialize(struct str *s, struct msg_unpacker *r) {"
print "\tuint32_t len = 0;"
print "\tif (!msg_unpacker_u32(r, &len))"
print "\t\treturn false;"
print "\tif (msg_unpacker_get_available(r) < len)"
print "\t\treturn false;"
print "\t*s = str_make();"
print "\tstr_append_data(s, r->data + r->offset, len);"
print "\tr->offset += len;"
print "\tif (!utf8_validate (s->str, s->len))"
print "\t\treturn false;"
print "\treturn true;"
print "}"
}
function codegen_constant(name, value) {
print ""
print "enum { " PrefixUpper name " = " value " };"
}
function codegen_enum_value(name, subname, value, cg) {
append(cg, "fields",
"\t" PrefixUpper toupper(cameltosnake(name)) "_" subname \
" = " value ",\n")
}
function codegen_enum(name, cg, ctype) {
ctype = "enum " PrefixLower cameltosnake(name)
print ""
print ctype " {"
print cg["fields"] "};"
# XXX: This should also check if it isn't out-of-range for any reason,
# but our usage of sprintf() stands in the way a bit.
CodegenSerialize[name] = "\tstr_pack_i8(w, %s);\n"
CodegenDeserialize[name] = \
"\t{\n" \
"\t\tint8_t v = 0;\n" \
"\t\tif (!msg_unpacker_i8(r, &v) || !v)\n" \
"\t\t\treturn false;\n" \
"\t\t%s = v;\n" \
"\t}\n"
CodegenCType[name] = ctype
for (i in cg)
delete cg[i]
}
function codegen_struct_tag(d, cg, f) {
f = "self->" d["name"]
append(cg, "fields", "\t" CodegenCType[d["type"]] " " d["name"] ";\n")
append(cg, "dispose", sprintf(CodegenDispose[d["type"]], f))
append(cg, "serialize", sprintf(CodegenSerialize[d["type"]], f))
# Do not deserialize here, that would be out of order.
}
function codegen_struct_field(d, cg, f, dispose, serialize, deserialize) {
f = "self->" d["name"]
dispose = CodegenDispose[d["type"]]
serialize = CodegenSerialize[d["type"]]
deserialize = CodegenDeserialize[d["type"]]
if (!d["isarray"]) {
append(cg, "fields", "\t" CodegenCType[d["type"]] " " d["name"] ";\n")
append(cg, "dispose", sprintf(dispose, f))
append(cg, "serialize", sprintf(serialize, f))
append(cg, "deserialize", sprintf(deserialize, f))
return
}
append(cg, "fields",
"\t" CodegenCType["u32"] " " d["name"] "_len;\n" \
"\t" CodegenCType[d["type"]] " *" d["name"] ";\n")
if (dispose)
append(cg, "dispose", "\tif (" f ")\n" \
"\t\tfor (size_t i = 0; i < " f "_len; i++)\n" \
indent(indent(sprintf(dispose, f "[i]"))))
append(cg, "dispose", "\tfree(" f ");\n")
append(cg, "serialize", sprintf(CodegenSerialize["u32"], f "_len"))
if (d["type"] == "u8" || d["type"] == "i8") {
append(cg, "serialize",
"\tstr_append_data(w, " f ", " f "_len);\n")
} else if (serialize) {
append(cg, "serialize",
"\tfor (size_t i = 0; i < " f "_len; i++)\n" \
indent(sprintf(serialize, f "[i]")))
}
append(cg, "deserialize", sprintf(CodegenDeserialize["u32"], f "_len") \
"\tif (!(" f " = calloc(" f "_len + 1, sizeof *" f ")))\n" \
"\t\treturn false;\n")
if (d["type"] == "u8" || d["type"] == "i8") {
append(cg, "deserialize",
"\tif (msg_unpacker_get_available(r) < " f "_len)\n" \
"\t\treturn false;\n" \
"\tmemcpy(" f ", r->data + r->offset, " f "_len);\n" \
"\tr->offset += " f "_len;\n")
} else if (deserialize) {
append(cg, "deserialize",
"\tfor (size_t i = 0; i < " f "_len; i++)\n" \
indent(sprintf(deserialize, f "[i]")))
}
}
function codegen_struct(name, cg, ctype, funcname) {
ctype = "struct " PrefixLower cameltosnake(name)
print ""
print ctype " {"
print cg["fields"] "};"
if (cg["dispose"]) {
funcname = PrefixLower cameltosnake(name) "_free"
print ""
print "static void\n" funcname "(" ctype " *self) {"
print cg["dispose"] "}"
CodegenDispose[name] = "\t" funcname "(&%s);\n"
}
if (cg["serialize"]) {
funcname = PrefixLower cameltosnake(name) "_serialize"
print ""
print "static bool\n" \
funcname "(\n\t\t" ctype " *self, struct str *w) {"
print cg["serialize"] "\treturn true;"
print "}"
CodegenSerialize[name] = "\tif (!" funcname "(&%s, w))\n" \
"\t\treturn false;\n"
}
if (cg["deserialize"]) {
funcname = PrefixLower cameltosnake(name) "_deserialize"
print ""
print "static bool\n" \
funcname "(\n\t\t" ctype " *self, struct msg_unpacker *r) {"
print cg["deserialize"] "\treturn true;"
print "}"
CodegenDeserialize[name] = "\tif (!" funcname "(&%s, r))\n" \
"\t\treturn false;\n"
}
CodegenCType[name] = ctype
for (i in cg)
delete cg[i]
}
function codegen_union_tag(d, cg) {
cg["tagtype"] = d["type"]
cg["tagname"] = d["name"]
append(cg, "fields", "\t" CodegenCType[d["type"]] " " d["name"] ";\n")
}
function codegen_union_struct( \
name, casename, cg, scg, structname, fieldname, fullcasename) {
# Don't generate obviously useless structs.
fullcasename = toupper(cameltosnake(cg["tagtype"])) "_" casename
if (!scg["dispose"] && !scg["deserialize"]) {
append(cg, "structless", "\tcase " PrefixUpper fullcasename ":\n")
for (i in scg)
delete scg[i]
return
}
# And thus not all generated structs are present in Types.
structname = name "_" casename
fieldname = tolower(casename)
codegen_struct(structname, scg)
append(cg, "fields", "\t" CodegenCType[structname] " " fieldname ";\n")
if (CodegenDispose[structname])
append(cg, "dispose", "\tcase " PrefixUpper fullcasename ":\n" \
indent(sprintf(CodegenDispose[structname], "self->" fieldname)) \
"\t\tbreak;\n")
# With no de/serialization code, this will simply recognize the tag.
append(cg, "serialize", "\tcase " PrefixUpper fullcasename ":\n" \
indent(sprintf(CodegenSerialize[structname], "self->" fieldname)) \
"\t\tbreak;\n")
append(cg, "deserialize", "\tcase " PrefixUpper fullcasename ":\n" \
indent(sprintf(CodegenDeserialize[structname], "self->" fieldname)) \
"\t\tbreak;\n")
}
function codegen_union(name, cg, f, ctype, funcname) {
ctype = "union " PrefixLower cameltosnake(name)
print ""
print ctype " {"
print cg["fields"] "};"
f = "self->" cg["tagname"]
if (cg["dispose"]) {
funcname = PrefixLower cameltosnake(name) "_free"
print ""
print "static void\n" funcname "(" ctype " *self) {"
print "\tswitch (" f ") {"
if (cg["structless"])
print cg["structless"] \
indent(sprintf(CodegenDispose[cg["tagtype"]], f)) "\t\tbreak;"
print cg["dispose"] "\tdefault:"
print "\t\tbreak;"
print "\t}"
print "}"
CodegenDispose[name] = "\t" funcname "(&%s);\n"
}
if (cg["serialize"]) {
funcname = PrefixLower cameltosnake(name) "_serialize"
print ""
print "static bool\n" \
funcname "(\n\t\t" ctype " *self, struct str *w) {"
print "\tswitch (" f ") {"
if (cg["structless"])
print cg["structless"] \
indent(sprintf(CodegenSerialize[cg["tagtype"]], f)) "\t\tbreak;"
print cg["serialize"] "\tdefault:"
print "\t\treturn false;"
print "\t}"
print "\treturn true;"
print "}"
CodegenSerialize[name] = "\tif (!" funcname "(&%s, w))\n" \
"\t\treturn false;\n"
}
if (cg["deserialize"]) {
funcname = PrefixLower cameltosnake(name) "_deserialize"
print ""
print "static bool\n" \
funcname "(\n\t\t" ctype " *self, struct msg_unpacker *r) {"
print sprintf(CodegenDeserialize[cg["tagtype"]], f)
print "\tswitch (" f ") {"
if (cg["structless"])
print cg["structless"] "\t\tbreak;"
print cg["deserialize"] "\tdefault:"
print "\t\treturn false;"
print "\t}"
print "\treturn true;"
print "}"
CodegenDeserialize[name] = "\tif (!" funcname "(&%s, r))\n" \
"\t\treturn false;\n"
}
CodegenCType[name] = ctype
for (i in cg)
delete cg[i]
}

View File

@ -1,538 +0,0 @@
# xC-gen-proto-go.awk: Go backend for xC-gen-proto.awk.
#
# Copyright (c) 2022, Přemysl Eric Janouch <p@janouch.name>
# SPDX-License-Identifier: 0BSD
#
# This backend also enables proxying to other endpoints using JSON.
function define_internal(name, gotype) {
Types[name] = "internal"
CodegenGoType[name] = gotype
}
function define_sint(size, shortname, gotype) {
shortname = "i" size
gotype = "int" size
define_internal(shortname, gotype)
CodegenAppendJSON[shortname] = \
"\tb = strconv.AppendInt(b, int64(%s), 10)\n"
if (size == 8) {
CodegenSerialize[shortname] = "\tdata = append(data, uint8(%s))\n"
CodegenDeserialize[shortname] = \
"\tif len(data) >= 1 {\n" \
"\t\t%s, data = int8(data[0]), data[1:]\n" \
"\t} else {\n" \
"\t\treturn nil, false\n" \
"\t}\n"
return
}
CodegenSerialize[shortname] = \
"\tdata = binary.BigEndian.AppendUint" size "(data, uint" size "(%s))\n"
CodegenDeserialize[shortname] = \
"\tif len(data) >= " (size / 8) " {\n" \
"\t\t%s = " gotype "(binary.BigEndian.Uint" size "(data))\n" \
"\t\tdata = data[" (size / 8) ":]\n" \
"\t} else {\n" \
"\t\treturn nil, false\n" \
"\t}\n"
}
function define_uint(size, shortname, gotype) {
# Both []byte and []uint8 luckily marshal as base64-encoded JSON strings,
# so there's no need to rename the type as an exception.
shortname = "u" size
gotype = "uint" size
define_internal(shortname, gotype)
CodegenAppendJSON[shortname] = \
"\tb = strconv.AppendUint(b, uint64(%s), 10)\n"
if (size == 8) {
CodegenSerialize[shortname] = "\tdata = append(data, %s)\n"
CodegenDeserialize[shortname] = \
"\tif len(data) >= 1 {\n" \
"\t\t%s, data = data[0], data[1:]\n" \
"\t} else {\n" \
"\t\treturn nil, false\n" \
"\t}\n"
return
}
CodegenSerialize[shortname] = \
"\tdata = binary.BigEndian.AppendUint" size "(data, %s)\n"
CodegenDeserialize[shortname] = \
"\tif len(data) >= " (size / 8) " {\n" \
"\t\t%s = binary.BigEndian.Uint" size "(data)\n" \
"\t\tdata = data[" (size / 8) ":]\n" \
"\t} else {\n" \
"\t\treturn nil, false\n" \
"\t}\n"
}
# Currently two outputs cannot coexist within the same package.
function codegen_private(name) {
return "proto" name
}
function codegen_begin( funcname) {
define_sint("8")
define_sint("16")
define_sint("32")
define_sint("64")
define_uint("8")
define_uint("16")
define_uint("32")
define_uint("64")
define_internal("bool", "bool")
define_internal("string", "string")
# Cater to "go generate", for what it's worth.
CodegenPackage = ENV["GOPACKAGE"]
if (!CodegenPackage)
CodegenPackage = "main"
print "// Code generated from " FILENAME ". DO NOT EDIT."
print ""
print "package " CodegenPackage
print ""
print "import ("
print "\t`encoding/base64`"
print "\t`encoding/binary`"
print "\t`encoding/json`"
print "\t`errors`"
print "\t`math`"
print "\t`strconv`"
print "\t`unicode/utf8`"
print ")"
print ""
CodegenAppendJSON["bool"] = \
"\tb = strconv.AppendBool(b, %s)\n"
CodegenSerialize["bool"] = \
"\tif %s {\n" \
"\t\tdata = append(data, 1)\n" \
"\t} else {\n" \
"\t\tdata = append(data, 0)\n" \
"\t}\n"
funcname = codegen_private("ConsumeBoolFrom")
print "// " funcname " tries to deserialize a boolean value"
print "// from the beginning of a byte stream. When successful,"
print "// it returns a subslice with any data that might follow."
print "func " funcname "(data []byte, b *bool) ([]byte, bool) {"
print "\tif len(data) < 1 {"
print "\t\treturn nil, false"
print "\t}"
print "\tif data[0] != 0 {"
print "\t\t*b = true"
print "\t} else {"
print "\t\t*b = false"
print "\t}"
print "\treturn data[1:], true"
print "}"
print ""
CodegenDeserialize["bool"] = \
"\tif data, ok = " funcname "(data, &%s); !ok {\n" \
"\t\treturn nil, ok\n" \
"\t}\n"
funcname = codegen_private("AppendStringTo")
print "// " funcname " tries to serialize a string value,"
print "// appending it to the end of a byte stream."
print "func " funcname "(data []byte, s string) ([]byte, bool) {"
print "\tif len(s) > math.MaxUint32 {"
print "\t\treturn nil, false"
print "\t}"
print "\tdata = binary.BigEndian.AppendUint32(data, uint32(len(s)))"
print "\treturn append(data, s...), true"
print "}"
print ""
CodegenSerialize["string"] = \
"\tif data, ok = " funcname "(data, %s); !ok {\n" \
"\t\treturn nil, ok\n" \
"\t}\n"
funcname = codegen_private("ConsumeStringFrom")
print "// " funcname " tries to deserialize a string value"
print "// from the beginning of a byte stream. When successful,"
print "// it returns a subslice with any data that might follow."
print "func " funcname "(data []byte, s *string) ([]byte, bool) {"
print "\tif len(data) < 4 {"
print "\t\treturn nil, false"
print "\t}"
print "\tlength := binary.BigEndian.Uint32(data)"
print "\tif data = data[4:]; uint64(len(data)) < uint64(length) {"
print "\t\treturn nil, false"
print "\t}"
print "\t*s = string(data[:length])"
print "\tif !utf8.ValidString(*s) {"
print "\t\treturn nil, false"
print "\t}"
print "\treturn data[length:], true"
print "}"
print ""
CodegenDeserialize["string"] = \
"\tif data, ok = " funcname "(data, &%s); !ok {\n" \
"\t\treturn nil, ok\n" \
"\t}\n"
funcname = codegen_private("UnmarshalEnumJSON")
print "// " funcname " converts a JSON fragment to an integer,"
print "// ensuring that it's within the expected range of enum values."
print "func " funcname "(data []byte) (int64, error) {"
print "\tvar n int64"
print "\tif err := json.Unmarshal(data, &n); err != nil {"
print "\t\treturn 0, err"
print "\t} else if n > math.MaxInt8 || n < math.MinInt8 {"
print "\t\treturn 0, errors.New(`integer out of range`)"
print "\t} else {"
print "\t\treturn n, nil"
print "\t}"
print "}"
print ""
}
function codegen_constant(name, value) {
print "const " PrefixCamel snaketocamel(name) " = " value
print ""
}
function codegen_enum_value(name, subname, value, cg, goname) {
goname = PrefixCamel name snaketocamel(subname)
append(cg, "fields",
"\t" goname " = " value "\n")
append(cg, "stringer",
"\tcase " goname ":\n" \
"\t\treturn `" snaketocamel(subname) "`\n")
append(cg, "marshal",
goname ",\n")
append(cg, "unmarshal",
"\tcase `" snaketocamel(subname) "`:\n" \
"\t\t*v = " goname "\n")
}
function codegen_enum(name, cg, gotype, fields, funcname) {
gotype = PrefixCamel name
print "type " gotype " int8"
print ""
print "const ("
print cg["fields"] ")"
print ""
print "func (v " gotype ") String() string {"
print "\tswitch v {"
print cg["stringer"] "\tdefault:"
print "\t\treturn strconv.Itoa(int(v))"
print "\t}"
print "}"
print ""
CodegenIsMarshaler[name] = 1
fields = cg["marshal"]
sub(/,\n$/, ":", fields)
gsub(/\n/, "\n\t", fields)
print "func (v " gotype ") MarshalJSON() ([]byte, error) {"
print "\tswitch v {"
print indent("case " fields)
print "\t\treturn []byte(`\"` + v.String() + `\"`), nil"
print "\t}"
print "\treturn json.Marshal(int(v))"
print "}"
print ""
funcname = codegen_private("UnmarshalEnumJSON")
print "func (v *" gotype ") UnmarshalJSON(data []byte) error {"
print "\tvar s string"
print "\tif json.Unmarshal(data, &s) == nil {"
print "\t\t// Handled below."
print "\t} else if n, err := " funcname "(data); err != nil {"
print "\t\treturn err"
print "\t} else {"
print "\t\t*v = " gotype "(n)"
print "\t\treturn nil"
print "\t}"
print ""
print "\tswitch s {"
print cg["unmarshal"] "\tdefault:"
print "\t\treturn errors.New(`unrecognized value: ` + s)"
print "\t}"
print "\treturn nil"
print "}"
print ""
# XXX: This should also check if it isn't out-of-range for any reason,
# but our usage of sprintf() stands in the way a bit.
CodegenSerialize[name] = "\tdata = append(data, uint8(%s))\n"
CodegenDeserialize[name] = \
"\tif len(data) >= 1 {\n" \
"\t\t%s, data = " gotype "(data[0]), data[1:]\n" \
"\t} else {\n" \
"\t\treturn nil, false\n" \
"\t}\n"
CodegenGoType[name] = gotype
for (i in cg)
delete cg[i]
}
function codegen_marshal(type, f, marshal) {
if (CodegenAppendJSON[type])
return sprintf(CodegenAppendJSON[type], f)
# Complex types are json.Marshalers, there's no need to json.Marshal(&f).
if (CodegenIsMarshaler[type])
marshal = f ".MarshalJSON()"
else
marshal = "json.Marshal(" f ")"
return \
"\tif j, err := " marshal "; err != nil {\n" \
"\t\treturn nil, err\n" \
"\t} else {\n" \
"\t\tb = append(b, j...)\n" \
"\t}\n"
}
function codegen_struct_field_marshal(d, cg, camel, f, marshal) {
camel = snaketocamel(d["name"])
f = "s." camel
if (!d["isarray"]) {
append(cg, "marshal",
"\tb = append(b, `,\"" decapitalize(camel) "\":`...)\n" \
codegen_marshal(d["type"], f))
return
}
# Note that we do not produce `null` for nil slices, unlike encoding/json.
# And arrays never get deserialized as such.
if (d["type"] == "u8") {
append(cg, "marshal",
"\tb = append(b, `,\"" decapitalize(camel) "\":\"`...)\n" \
"\tb = append(b, base64.StdEncoding.EncodeToString(" f ")...)\n" \
"\tb = append(b, '\"')\n")
return
}
append(cg, "marshal",
"\tb = append(b, `,\"" decapitalize(camel) "\":[`...)\n" \
"\tfor i := 0; i < len(" f "); i++ {\n" \
"\t\tif i > 0 {\n" \
"\t\t\tb = append(b, ',')\n" \
"\t\t}\n" \
indent(codegen_marshal(d["type"], f "[i]")) \
"\t}\n" \
"\tb = append(b, ']')\n")
}
function codegen_struct_field(d, cg, camel, f, serialize, deserialize) {
codegen_struct_field_marshal(d, cg)
camel = snaketocamel(d["name"])
f = "s." camel
serialize = CodegenSerialize[d["type"]]
deserialize = CodegenDeserialize[d["type"]]
if (!d["isarray"]) {
append(cg, "fields", "\t" camel " " CodegenGoType[d["type"]] \
" `json:\"" decapitalize(camel) "\"`\n")
append(cg, "serialize", sprintf(serialize, f))
append(cg, "deserialize", sprintf(deserialize, f))
return
}
append(cg, "fields", "\t" camel " []" CodegenGoType[d["type"]] \
" `json:\"" decapitalize(camel) "\"`\n")
# XXX: This should also check if it isn't out-of-range for any reason.
append(cg, "serialize",
sprintf(CodegenSerialize["u32"], "uint32(len(" f "))"))
if (d["type"] == "u8") {
append(cg, "serialize",
"\tdata = append(data, " f "...)\n")
} else {
append(cg, "serialize",
"\tfor i := 0; i < len(" f "); i++ {\n" \
indent(sprintf(serialize, f "[i]")) \
"\t}\n")
}
append(cg, "deserialize",
"\t{\n" \
"\t\tvar length uint32\n" \
indent(sprintf(CodegenDeserialize["u32"], "length")))
if (d["type"] == "u8") {
append(cg, "deserialize",
"\t\tif uint64(len(data)) < uint64(length) {\n" \
"\t\t\treturn nil, false\n" \
"\t\t}\n" \
"\t\t" f ", data = data[:length], data[length:]\n" \
"\t}\n")
} else {
append(cg, "deserialize",
"\t\t" f " = make([]" CodegenGoType[d["type"]] ", length)\n" \
"\t}\n" \
"\tfor i := 0; i < len(" f "); i++ {\n" \
indent(sprintf(deserialize, f "[i]")) \
"\t}\n")
}
}
function codegen_struct_tag(d, cg, camel, f) {
codegen_struct_field_marshal(d, cg)
camel = snaketocamel(d["name"])
f = "s." camel
append(cg, "fields", "\t" camel " " CodegenGoType[d["type"]] \
" `json:\"" decapitalize(camel) "\"`\n")
append(cg, "serialize", sprintf(CodegenSerialize[d["type"]], f))
# Do not deserialize here, that is already done by the containing union.
}
function codegen_struct(name, cg, gotype) {
gotype = PrefixCamel name
print "type " gotype " struct {\n" cg["fields"] "}\n"
if (cg["marshal"]) {
CodegenIsMarshaler[name] = 1
print "func (s *" gotype ") MarshalJSON() ([]byte, error) {"
print "\tb := []byte{}"
print cg["marshal"] "\tb[0] = '{'"
print "\treturn append(b, '}'), nil"
print "}"
print ""
}
if (cg["serialize"]) {
print "func (s *" gotype ") AppendTo(data []byte) ([]byte, bool) {"
print "\tok := true"
print cg["serialize"] "\treturn data, ok"
print "}"
print ""
CodegenSerialize[name] = \
"\tif data, ok = %s.AppendTo(data); !ok {\n" \
"\t\treturn nil, ok\n" \
"\t}\n"
}
if (cg["deserialize"]) {
print "func (s *" gotype ") ConsumeFrom(data []byte) ([]byte, bool) {"
print "\tok := true"
print cg["deserialize"] "\treturn data, ok"
print "}"
print ""
CodegenDeserialize[name] = \
"\tif data, ok = %s.ConsumeFrom(data); !ok {\n" \
"\t\treturn nil, ok\n" \
"\t}\n"
}
CodegenGoType[name] = gotype
for (i in cg)
delete cg[i]
}
function codegen_union_tag(d, cg) {
cg["tagtype"] = d["type"]
cg["tagname"] = d["name"]
# The tag is implied from the type of struct stored in the interface.
}
function codegen_union_struct(name, casename, cg, scg, structname, init) {
# And thus not all generated structs are present in Types.
structname = name snaketocamel(casename)
codegen_struct(structname, scg)
init = CodegenGoType[structname] "{" snaketocamel(cg["tagname"]) \
": " decapitalize(snaketocamel(cg["tagname"])) "}"
append(cg, "unmarshal",
"\tcase " CodegenGoType[cg["tagtype"]] snaketocamel(casename) ":\n" \
"\t\ts := " init "\n" \
"\t\terr = json.Unmarshal(data, &s)\n" \
"\t\tu.Interface = &s\n")
append(cg, "serialize",
"\tcase *" CodegenGoType[structname] ":\n" \
indent(sprintf(CodegenSerialize[structname], "union")))
append(cg, "deserialize",
"\tcase " CodegenGoType[cg["tagtype"]] snaketocamel(casename) ":\n" \
"\t\ts := " init "\n" \
indent(sprintf(CodegenDeserialize[structname], "s")) \
"\t\tu.Interface = &s\n")
}
function codegen_union(name, cg, gotype, tagfield, tagvar) {
gotype = PrefixCamel name
print "type " gotype " struct {"
print "\tInterface any"
print "}"
print ""
# This cannot be a pointer method, it wouldn't work recursively.
CodegenIsMarshaler[name] = 1
print "func (u " gotype ") MarshalJSON() ([]byte, error) {"
print "\treturn u.Interface.(json.Marshaler).MarshalJSON()"
print "}"
print ""
tagfield = snaketocamel(cg["tagname"])
tagvar = decapitalize(tagfield)
print "func (u *" gotype ") UnmarshalJSON(data []byte) (err error) {"
print "\tvar t struct {"
print "\t\t" tagfield " " CodegenGoType[cg["tagtype"]] \
" `json:\"" tagvar "\"`"
print "\t}"
print "\tif err := json.Unmarshal(data, &t); err != nil {"
print "\t\treturn err"
print "\t}"
print ""
print "\tswitch " tagvar " := t." tagfield "; " tagvar " {"
print cg["unmarshal"] "\tdefault:"
print "\t\terr = errors.New(`unsupported value: ` + " tagvar ".String())"
print "\t}"
print "\treturn err"
print "}"
print ""
# XXX: Consider changing the interface into an AppendTo/ConsumeFrom one,
# that would eliminate these type case switches entirely.
# On the other hand, it would make it possible to send unsuitable structs.
print "func (u *" gotype ") AppendTo(data []byte) ([]byte, bool) {"
print "\tok := true"
print "\tswitch union := u.Interface.(type) {"
print cg["serialize"] "\tdefault:"
print "\t\treturn nil, false"
print "\t}"
print "\treturn data, ok"
print "}"
print ""
CodegenSerialize[name] = \
"\tif data, ok = %s.AppendTo(data); !ok {\n" \
"\t\treturn nil, ok\n" \
"\t}\n"
print "func (u *" gotype ") ConsumeFrom(data []byte) ([]byte, bool) {"
print "\tok := true"
print "\tvar " tagvar " " CodegenGoType[cg["tagtype"]]
print sprintf(CodegenDeserialize[cg["tagtype"]], tagvar)
print "\tswitch " tagvar " {"
print cg["deserialize"] "\tdefault:"
print "\t\treturn nil, false"
print "\t}"
print "\treturn data, ok"
print "}"
print ""
CodegenDeserialize[name] = \
"\tif data, ok = %s.ConsumeFrom(data); !ok {\n" \
"\t\treturn nil, ok\n" \
"\t}\n"
CodegenGoType[name] = gotype
for (i in cg)
delete cg[i]
}

View File

@ -1,225 +0,0 @@
# xC-gen-proto-js.awk: Javascript backend for xC-gen-proto.awk.
#
# Copyright (c) 2022, Přemysl Eric Janouch <p@janouch.name>
# SPDX-License-Identifier: 0BSD
#
# This backend is currently for decoding the binary format only.
# (JSON is way too expensive to process and transfer.)
#
# Import the resulting script as a Javascript module.
function define_internal(name) {
Types[name] = "internal"
}
function define_sint(size, shortname) {
shortname = "i" size
define_internal(shortname)
CodegenDeserialize[shortname] = "\t%s = r." shortname "()\n"
print ""
print "\t" shortname "() {"
if (size == "64") {
# XXX: 2^53 - 1 must be enough for anyone. BigInts are a PITA.
print "\t\tconst " shortname \
" = Number(this.getBigInt" size "(this.offset))"
} else {
print "\t\tconst " shortname " = this.getInt" size "(this.offset)"
}
print "\t\tthis.offset += " (size / 8)
print "\t\treturn " shortname
print "\t}"
}
function define_uint(size, shortname) {
shortname = "u" size
define_internal(shortname)
CodegenDeserialize[shortname] = "\t%s = r." shortname "()\n"
print ""
print "\t" shortname "() {"
if (size == "64") {
# XXX: 2^53 - 1 must be enough for anyone. BigInts are a PITA.
print "\t\tconst " shortname \
" = Number(this.getBigUint" size "(this.offset))"
} else {
print "\t\tconst " shortname " = this.getUint" size "(this.offset)"
}
print "\t\tthis.offset += " (size / 8)
print "\t\treturn " shortname
print "\t}"
}
function codegen_begin() {
print "// Code generated from " FILENAME ". DO NOT EDIT."
print ""
print "export class Reader extends DataView {"
print "\tconstructor() {"
print "\t\tsuper(...arguments)"
print "\t\tthis.offset = 0"
print "\t\tthis.decoder = new TextDecoder('utf-8', {fatal: true})"
print "\t}"
print ""
print "\tget empty() {"
print "\t\treturn this.byteLength <= this.offset"
print "\t}"
print ""
print "\trequire(len) {"
print "\t\tif (this.byteLength - this.offset < len)"
print "\t\t\tthrow `Premature end of data`"
print "\t\treturn this.byteOffset + this.offset"
print "\t}"
define_internal("string")
CodegenDeserialize["string"] = "\t%s = r.string()\n"
print ""
print "\tstring() {"
print "\t\tconst len = this.getUint32(this.offset)"
print "\t\tthis.offset += 4"
print "\t\tconst array = new Uint8Array("
print "\t\t\tthis.buffer, this.require(len), len)"
print "\t\tthis.offset += len"
print "\t\treturn this.decoder.decode(array)"
print "\t}"
define_internal("bool")
CodegenDeserialize["bool"] = "\t%s = r.bool()\n"
print ""
print "\tbool() {"
print "\t\tconst u8 = this.getUint8(this.offset)"
print "\t\tthis.offset += 1"
print "\t\treturn u8 != 0"
print "\t}"
define_sint("8")
define_sint("16")
define_sint("32")
define_sint("64")
define_uint("8")
define_uint("16")
define_uint("32")
define_uint("64")
print "}"
}
function codegen_constant(name, value) {
print ""
print "export const " decapitalize(snaketocamel(name)) " = " value
}
function codegen_enum_value(name, subname, value, cg) {
append(cg, "fields", "\t" snaketocamel(subname) ": " value ",\n")
}
function codegen_enum(name, cg) {
print ""
print "export const " name " = Object.freeze({"
print cg["fields"] "})"
CodegenDeserialize[name] = "\t%s = r.i8()\n"
for (i in cg)
delete cg[i]
}
function codegen_struct_field(d, cg, camel, f, deserialize) {
camel = decapitalize(snaketocamel(d["name"]))
f = "s." camel
append(cg, "fields", "\t" camel "\n")
deserialize = CodegenDeserialize[d["type"]]
if (!d["isarray"]) {
append(cg, "deserialize", sprintf(deserialize, f))
return
}
append(cg, "deserialize",
"\t{\n" \
indent(sprintf(CodegenDeserialize["u32"], "const len")))
if (d["type"] == "u8") {
append(cg, "deserialize",
"\t\t" f " = new Uint8Array(\n" \
"\t\t\tr.buffer, r.require(len), len)\n" \
"\t\tr.offset += len\n" \
"\t}\n")
return
}
if (d["type"] == "i8") {
append(cg, "deserialize",
"\t\t" f " = new Int8Array(\n" \
"\t\t\tr.buffer, r.require(len), len)\n" \
"\t\tr.offset += len\n" \
"\t}\n")
return
}
append(cg, "deserialize",
"\t\t" f " = new Array(len)\n" \
"\t}\n" \
"\tfor (let i = 0; i < " f ".length; i++)\n" \
indent(sprintf(deserialize, f "[i]")))
}
function codegen_struct_tag(d, cg) {
append(cg, "fields", "\t" decapitalize(snaketocamel(d["name"])) "\n")
# Do not deserialize here, that is already done by the containing union.
}
function codegen_struct(name, cg) {
print ""
print "export class " name " {"
print cg["fields"] cg["methods"]
print "\tstatic deserialize(r) {"
print "\t\tconst s = new " name "()"
print indent(cg["deserialize"]) "\t\treturn s"
print "\t}"
print "}"
CodegenDeserialize[name] = "\t%s = " name ".deserialize(r)\n"
for (i in cg)
delete cg[i]
}
function codegen_union_tag(d, cg) {
cg["tagtype"] = d["type"]
cg["tagname"] = d["name"]
}
function codegen_union_struct(name, casename, cg, scg, structname) {
append(scg, "methods",
"\n" \
"\tconstructor() {\n" \
"\t\tthis." decapitalize(snaketocamel(cg["tagname"])) \
" = " cg["tagtype"] "." snaketocamel(casename) "\n" \
"\t}\n")
# And thus not all generated structs are present in Types.
structname = name snaketocamel(casename)
codegen_struct(structname, scg)
append(cg, "deserialize",
"\tcase " cg["tagtype"] "." snaketocamel(casename) ":\n" \
"\t{\n" \
indent(sprintf(CodegenDeserialize[structname], "const s")) \
"\t\treturn s\n" \
"\t}\n")
}
function codegen_union(name, cg, tagvar) {
tagvar = decapitalize(snaketocamel(cg["tagname"]))
print ""
print "export function deserialize" name "(r) {"
print sprintf(CodegenDeserialize[cg["tagtype"]], "const " tagvar) \
"\tswitch (" tagvar ") {"
print cg["deserialize"] "\tdefault:"
print "\t\tthrow `Unknown " cg["tagtype"] " (${tagvar})`"
print "\t}"
print "}"
CodegenDeserialize[name] = "\t%s = deserialize" name "(r)\n"
for (i in cg)
delete cg[i]
}

View File

@ -1,307 +0,0 @@
# xC-gen-proto.awk: an XDR-derived code generator for network protocols.
#
# Copyright (c) 2022, Přemysl Eric Janouch <p@janouch.name>
# SPDX-License-Identifier: 0BSD
#
# You may read RFC 4506 for context, however it is only a source of inspiration.
# Grammar is easy to deduce from the parser.
#
# Native types: bool, u{8,16,32,64}, i{8,16,32,64}, string
#
# Don't define any new types, unless you hate yourself, then it's okay to do so.
# Backends tend to be a pain in the arse, for different reasons.
#
# All numbers are encoded in big-endian byte order.
# Booleans are one byte each.
# Strings must be valid UTF-8, use u8<> to lift that restriction.
# String and array lengths are encoded as u32.
# Enumeration values automatically start at 1, and are encoded as i8.
# Any struct or union field may be a variable-length array.
#
# Message framing is done externally, but also happens to prefix u32 lengths,
# unless this role is already filled by, e.g., WebSocket.
#
# Usage: env LC_ALL=C awk -f xC-gen-proto.awk -f xC-gen-proto-{c,go,js}.awk \
# -v PrefixCamel=Relay xC-proto > xC-proto.{c,go,js} | {clang-format,gofmt,...}
# --- Utilities ----------------------------------------------------------------
function cameltosnake(s) {
while (match(s, /[[:lower:]][[:upper:]]/)) {
s = substr(s, 1, RSTART) "_" \
tolower(substr(s, RSTART + 1, RLENGTH - 1)) \
substr(s, RSTART + RLENGTH)
}
return tolower(s)
}
function snaketocamel(s) {
s = toupper(substr(s, 1, 1)) tolower(substr(s, 2))
while (match(s, /_[[:alnum:]]/)) {
s = substr(s, 1, RSTART - 1) \
toupper(substr(s, RSTART + 1, RLENGTH - 1)) \
substr(s, RSTART + RLENGTH)
}
return s
}
function decapitalize(s) {
if (match(s, /[[:upper:]][[:lower:]]/)) {
return tolower(substr(s, 1, 1)) substr(s, 2)
}
return s
}
function indent(s) {
if (!s)
return s
gsub(/\n/, "\n\t", s)
sub(/\t*$/, "", s)
return "\t" s
}
function append(a, key, value) {
a[key] = a[key] value
}
# --- Parsing ------------------------------------------------------------------
function fatal(message) {
print "// " FILENAME ":" FNR ": fatal error: " message
print FILENAME ":" FNR ": fatal error: " message > "/dev/stderr"
exit 1
}
function skipcomment() {
do {
if (match($0, /[*][/]/)) {
$0 = substr($0, RSTART + RLENGTH)
return
}
} while (getline > 0)
fatal("unterminated block comment")
}
function nexttoken() {
do {
if (match($0, /^[[:space:]]+/)) {
$0 = substr($0, RLENGTH + 1)
} else if (match($0, /^[/][/].*/)) {
$0 = ""
} else if (match($0, /^[/][*]/)) {
$0 = substr($0, RLENGTH + 1)
skipcomment()
} else if (match($0, /^[[:alpha:]][[:alnum:]_]*/)) {
Token = substr($0, 1, RLENGTH)
$0 = substr($0, RLENGTH + 1)
return Token
} else if (match($0, /^(0[xX][0-9a-fA-F]+|[1-9][0-9]*)/)) {
Token = substr($0, 1, RLENGTH)
$0 = substr($0, RLENGTH + 1)
return Token
} else if ($0) {
Token = substr($0, 1, 1)
$0 = substr($0, 2)
return Token
}
} while ($0 || getline > 0)
Token = ""
return Token
}
function expect(v) {
if (!v)
fatal("broken expectations at `" Token "' before `" $0 "'")
return v
}
function accept(what) {
if (Token != what)
return 0
nexttoken()
return 1
}
function identifier( v) {
if (Token !~ /^[[:alpha:]]/)
return 0
v = Token
nexttoken()
return v
}
function number( v) {
if (Token !~ /^[0-9]/)
return 0
v = Token
nexttoken()
return v
}
function readnumber( ident) {
ident = identifier()
if (!ident)
return expect(number())
if (!(ident in Consts))
fatal("unknown constant: " ident)
return Consts[ident]
}
function defconst( ident, num) {
if (!accept("const"))
return 0
ident = expect(identifier())
expect(accept("="))
num = readnumber()
if (ident in Consts)
fatal("constant redefined: " ident)
Consts[ident] = num
codegen_constant(ident, num)
return 1
}
function readtype( ident) {
ident = deftype()
if (ident)
return ident
ident = identifier()
if (!ident)
return 0
if (!(ident in Types))
fatal("unknown type: " ident)
return ident
}
function defenum( name, ident, value, cg) {
delete cg[0]
name = expect(identifier())
expect(accept("{"))
while (!accept("}")) {
ident = expect(identifier())
value = value + 1
if (accept("="))
value = readnumber()
if (!value)
fatal("enumeration values cannot be zero")
if (value < -128 || value > 127)
fatal("enumeration value out of range")
expect(accept(","))
append(EnumValues, name, SUBSEP ident)
if (EnumValues[name, ident]++)
fatal("duplicate enum value: " ident)
codegen_enum_value(name, ident, value, cg)
}
Types[name] = "enum"
codegen_enum(name, cg)
return name
}
function readfield(out, nonvoid) {
nonvoid = !accept("void")
if (nonvoid) {
out["type"] = expect(readtype())
out["name"] = expect(identifier())
# TODO: Consider supporting XDR's VLA length limits here.
# TODO: Consider supporting XDR's fixed-length syntax for string limits.
out["isarray"] = accept("<") && expect(accept(">"))
}
expect(accept(";"))
return nonvoid
}
function defstruct( name, d, cg) {
delete d[0]
delete cg[0]
name = expect(identifier())
expect(accept("{"))
while (!accept("}")) {
if (readfield(d))
codegen_struct_field(d, cg)
}
Types[name] = "struct"
codegen_struct(name, cg)
return name
}
function defunion( name, tag, tagtype, tagvalue, cg, scg, d, a, i, unseen) {
delete cg[0]
delete scg[0]
delete d[0]
name = expect(identifier())
expect(accept("switch"))
expect(accept("("))
tag["type"] = tagtype = expect(readtype())
tag["name"] = expect(identifier())
expect(accept(")"))
if (Types[tagtype] != "enum")
fatal("not an enum type: " tagtype)
codegen_union_tag(tag, cg)
split(EnumValues[tagtype], a, SUBSEP)
for (i in a)
unseen[a[i]]++
expect(accept("{"))
while (!accept("}")) {
if (accept("case")) {
if (tagvalue)
codegen_union_struct(name, tagvalue, cg, scg)
tagvalue = expect(identifier())
expect(accept(":"))
if (!unseen[tagvalue]--)
fatal("no such value or duplicate case: " tagtype "." tagvalue)
codegen_struct_tag(tag, scg)
} else if (tagvalue) {
if (readfield(d))
codegen_struct_field(d, scg)
} else {
fatal("union fields must fall under a case")
}
}
if (tagvalue)
codegen_union_struct(name, tagvalue, cg, scg)
# What remains non-zero in unseen[2..] is simply not recognized/allowed.
Types[name] = "union"
codegen_union(name, cg)
return name
}
function deftype() {
if (accept("enum"))
return defenum()
if (accept("struct"))
return defstruct()
if (accept("union"))
return defunion()
return 0
}
{
if (PrefixCamel) {
PrefixLower = tolower(cameltosnake(PrefixCamel)) "_"
PrefixUpper = toupper(cameltosnake(PrefixCamel)) "_"
}
# This is not in a BEGIN clause (even though it consumes all input),
# so that the code generator can insert the first FILENAME.
codegen_begin()
nexttoken()
while (Token != "") {
expect(defconst() || deftype())
expect(accept(";"))
}
}

View File

View File

@ -2,16 +2,17 @@
.SUFFIXES:
AWK = env LC_ALL=C awk
tools = ../liberty/tools
outputs = xP proto.go public/proto.js public/mithril.js
all: $(outputs) public/ircfmt.woff2
xP: xP.go proto.go
go build -o $@
proto.go: ../xC-gen-proto.awk ../xC-gen-proto-go.awk ../xC-proto
$(AWK) -f ../xC-gen-proto.awk -f ../xC-gen-proto-go.awk \
-v PrefixCamel=Relay ../xC-proto > $@
public/proto.js: ../xC-gen-proto.awk ../xC-gen-proto-js.awk ../xC-proto
$(AWK) -f ../xC-gen-proto.awk -f ../xC-gen-proto-js.awk ../xC-proto > $@
proto.go: $(tools)/lxdrgen.awk $(tools)/lxdrgen-go.awk ../xC.lxdr
$(AWK) -f $(tools)/lxdrgen.awk -f $(tools)/lxdrgen-go.awk \
-v PrefixCamel=Relay ../xC.lxdr > $@
public/proto.js: $(tools)/lxdrgen.awk $(tools)/lxdrgen-mjs.awk ../xC.lxdr
$(AWK) -f $(tools)/lxdrgen.awk -f $(tools)/lxdrgen-mjs.awk ../xC.lxdr > $@
public/ircfmt.woff2: gen-ircfmt.awk
$(AWK) -v Output=$@ -f gen-ircfmt.awk
public/mithril.js: