progress. still not working. this is incredibly difficult.

7 files changed, 347 insertions, 60 deletions

diff --git a/nexgb/xgbgen/context.go b/nexgb/xgbgen/context.go
index e5acb12..712cad4 100644
--- a/nexgb/xgbgen/context.go
+++ b/nexgb/xgbgen/context.go
@@ -5,7 +5,6 @@ import (
 	"encoding/xml"
 	"fmt"
 	"log"
-	"strings"
 )
 
 type Context struct {
@@ -33,31 +32,6 @@ func (c *Context) Put(format string, v ...interface{}) {
 	}
 }
 
-// TypePrefix searches the parsed XML for a type matching 'needle'.
-// It then returns the appropriate prefix to be used in source code.
-// Note that the core X protocol *is* a namespace, but does not have a prefix.
-// Also note that you should probably check the BaseTypeMap and TypeMap
-// before calling this function.
-func (c *Context) TypePrefix(needle Type) string {
-	// If this is xproto, quit. No prefixes needed.
-	if c.xml.Header == "xproto" {
-		return ""
-	}
-
-	// First check for the type in the current namespace.
-	if c.xml.HasType(needle) {
-		return strings.Title(c.xml.Header)
-	}
-
-	// Now check each of the imports...
-	for _, imp := range c.xml.Imports {
-		if imp.xml.Header != "xproto" && imp.xml.HasType(needle) {
-			return strings.Title(imp.xml.Header)
-		}
-	}
-
-	return ""
-}
 
 // Translate is the big daddy of them all. It takes in an XML byte slice
 // and writes Go code to the 'out' buffer.
@@ -76,14 +50,4 @@ func (c *Context) Translate(xmlBytes []byte) {
 
 	// It's Morphin' Time!
 	c.xml.Morph(c)
-
-	// for _, req := range c.xml.Requests { 
-		// if req.Name != "CreateContext" && req.Name != "MakeCurrent" { 
-			// continue 
-		// } 
-		// log.Println(req.Name) 
-		// for _, field := range req.Fields { 
-			// log.Println("\t", field.XMLName.Local, field.Type.Morph(c)) 
-		// } 
-	// } 
 }
diff --git a/nexgb/xgbgen/go.go b/nexgb/xgbgen/go.go
index eb3f0fb..bfc54dd 100644
--- a/nexgb/xgbgen/go.go
+++ b/nexgb/xgbgen/go.go
@@ -8,12 +8,22 @@ package main
 
 	* Imports and helper variables.
 	* Manual type and name override maps.
+	* Constants for tweaking various morphing functions.
 	* Helper morphing functions.
+	* Morphing functions for each "sub-unit."
 	* Morphing functions for each "unit".
 	* Morphing functions for collections of "units".
+	
+	Units can be thought of as the top-level elements in an XML protocol
+	description file. Namely, structs, xidtypes, imports, enums, unions, etc.
+	Collections of units are simply "all of the UNIT in the XML file."
+	Sub-units can be thought of as recurring bits like struct contents (which
+	is used in events, replies, requests, errors, etc.) and expression
+	evaluation.
 */
 
 import (
+	"log"
 	"strings"
 )
 
@@ -38,6 +48,24 @@ var BaseTypeMap = map[string]string{
 	"BOOL": "bool",
 	"float": "float64",
 	"double": "float64",
+	"char": "byte",
+}
+
+// BaseTypeSizes should have precisely the same keys as in BaseTypeMap,
+// and the values should correspond to the size of the type in bytes.
+var BaseTypeSizes = map[string]uint{
+	"CARD8": 1,
+	"CARD16": 2,
+	"CARD32": 4,
+	"INT8": 1,
+	"INT16": 2,
+	"INT32": 4,
+	"BYTE": 1,
+	"BOOL": 1,
+	"float": 4,
+	"double": 8,
+	"char": 1,
+	"Id": 4,
 }
 
 // TypeMap is a map from types in the XML to type names that is used
@@ -54,12 +82,26 @@ var TypeMap = map[string]string{
 var NameMap = map[string]string{ }
 
 /******************************************************************************/
-// Helper functions that aide in morphing repetive constructs.
-// i.e., "structure contents", expressions, type and identifier names, etc.
+// Constants for changing the semantics of morphing functions.
+// These are mainly used to tweaking the writing of fields.
+// Namely, reading/writing is not exactly the same across events,
+// requests/replies and errors.
+/******************************************************************************/
+const (
+	FieldsEvent = iota
+	FieldsRequestReply
+	FieldsError
+)
+
+/******************************************************************************/
+// Helper functions that aide in morphing repetitive constructs.
+// i.e., type and identifier names, etc.
 /******************************************************************************/
 
 // Morph changes every TYPE (not names) into something suitable
-// for your language.
+// for your language. It also handles adding suffixes like 'Event'
+// and 'Union'. (A 'Union' suffix is used in Go because unions aren't
+// supported at the language level.)
 func (typ Type) Morph(c *Context) string {
 	t := string(typ)
 
@@ -87,7 +129,33 @@ func (typ Type) Morph(c *Context) string {
 
 	// Since there is no namespace, we need to look for a namespace
 	// in the current context.
-	return c.TypePrefix(typ) + splitAndTitle(t)
+	return typ.Prefix(c) + splitAndTitle(t)
+}
+
+// Prefix searches the parsed XML for a type matching 'typ'.
+// It then returns the appropriate prefix to be used in source code.
+// Note that the core X protocol *is* a namespace, but does not have a prefix.
+// Also note that you should probably check the BaseTypeMap and TypeMap
+// before calling this function.
+func (typ Type) Prefix(c *Context) string {
+	// If this is xproto, quit. No prefixes needed.
+	if c.xml.Header == "xproto" {
+		return ""
+	}
+
+	// First check for the type in the current namespace.
+	if c.xml.HasType(typ) {
+		return strings.Title(c.xml.Header)
+	}
+
+	// Now check each of the imports...
+	for _, imp := range c.xml.Imports {
+		if imp.xml.Header != "xproto" && imp.xml.HasType(typ) {
+			return strings.Title(imp.xml.Header)
+		}
+	}
+
+	return ""
 }
 
 // Morph changes every identifier (NOT type) into something suitable
@@ -104,6 +172,147 @@ func (name Name) Morph(c *Context) string {
 }
 
 /******************************************************************************/
+// Sub-unit morphing.
+// Below are functions that morph sub-units. Like collections of fields,
+// expressions, etc.
+// Note that collections of fields can be used in three different contexts:
+// definitions, reading from the wire and writing to the wire. Thus, there
+// exists 'MorphDefine', 'MorphRead', 'MorphWrite' defined on Fields.
+/******************************************************************************/
+func (fields Fields) MorphDefine(c *Context) {
+	for _, field := range fields {
+		field.MorphDefine(c)
+	}
+}
+
+func (field *Field) MorphDefine(c *Context) {
+	// We omit 'pad' and 'exprfield'
+	switch field.XMLName.Local {
+	case "field":
+		c.Putln("%s %s", field.Name.Morph(c), field.Type.Morph(c))
+	case "list":
+		c.Putln("%s []%s", field.Name.Morph(c), field.Type.Morph(c))
+	case "localfield":
+		c.Putln("%s %s", field.Name.Morph(c), field.Type.Morph(c))
+	case "valueparam":
+		c.Putln("%s %s", field.ValueMaskName.Morph(c),
+			field.ValueMaskType.Morph(c))
+		c.Putln("%s []%s", field.ValueListName.Morph(c),
+			field.ValueMaskType.Morph(c))
+	case "switch":
+		field.Bitcases.MorphDefine(c)
+	}
+}
+
+func (bitcases Bitcases) MorphDefine(c *Context) {
+	for _, bitcase := range bitcases {
+		bitcase.MorphDefine(c)
+	}
+}
+
+func (bitcase *Bitcase) MorphDefine(c *Context) {
+	bitcase.Fields.MorphDefine(c)
+}
+
+func (fields Fields) MorphRead(c *Context, kind int, evNoSeq bool) {
+	var nextByte uint
+
+	switch kind {
+	case FieldsEvent:
+		nextByte = 1
+	}
+
+	for _, field := range fields {
+		nextByte = field.MorphRead(c, kind, nextByte)
+		switch kind {
+		case FieldsEvent:
+			// Skip the sequence id
+			if !evNoSeq && (nextByte == 2 || nextByte == 3) {
+				nextByte = 4
+			}
+		}
+	}
+}
+
+func (field *Field) MorphRead(c *Context, kind int, byt uint) uint {
+	consumed := uint(0)
+	switch field.XMLName.Local {
+	case "pad":
+		consumed = uint(field.Bytes)
+	case "field":
+		if field.Type == "ClientMessageData" {
+			break
+		}
+		size := field.Type.Size(c)
+		typ := field.Type.Morph(c)
+		name := field.Name.Morph(c)
+		_, isBase := BaseTypeMap[string(field.Type)]
+
+		c.Put("v.%s = ", name)
+		if !isBase {
+			c.Put("%s(", typ)
+		}
+		switch size {
+		case 1:	c.Put("buf[%d]", byt)
+		case 2: c.Put("get16(buf[%d:])", byt)
+		case 4: c.Put("get32(buf[%d:])", byt)
+		case 8: c.Put("get64(buf[%d:])", byt)
+		default:
+			log.Fatalf("Unsupported field size '%d' for field '%s'.",
+				size, field)
+		}
+		if !isBase {
+			c.Put(")")
+		}
+		c.Putln("")
+
+		consumed = size
+	case "list":
+		c.Putln("")
+	}
+	return byt + consumed
+}
+
+func (fields Fields) MorphWrite(c *Context, kind int) {
+	var nextByte uint
+
+	switch kind {
+	case FieldsEvent:
+		nextByte = 1
+	}
+
+	for _, field := range fields {
+		nextByte = field.MorphWrite(c, kind, nextByte)
+	}
+}
+
+func (field *Field) MorphWrite(c *Context, kind int, byt uint) uint {
+	consumed := uint(0)
+	switch field.XMLName.Local {
+	case "pad":
+		consumed = uint(field.Bytes)
+	case "field":
+		size := field.Type.Size(c)
+		typ := field.Type.Morph(c)
+		name := field.Name.Morph(c)
+		switch size {
+		case 1:
+			c.Putln("v.%s = %s(buf[%d])", name, typ, byt)
+		case 2:
+			c.Putln("v.%s = %s(get16(buf[%d:]))", name, typ, byt)
+		case 4:
+			c.Putln("v.%s = %s(get32(buf[%d:]))", name, typ, byt)
+		case 8:
+			c.Putln("v.%s = %s(get64(buf[%d:]))", name, typ, byt)
+		}
+		consumed = size
+	case "list":
+		c.Putln("IDK")
+	}
+	return byt + consumed
+}
+
+/******************************************************************************/
 // Per element morphing.
 // Below are functions that morph a single unit.
 /******************************************************************************/
@@ -133,15 +342,23 @@ func (xid *Xid) Morph(c *Context) {
 
 // TypeDef morphing.
 func (typedef *TypeDef) Morph(c *Context) {
-	c.Putln("type %s %s", typedef.Old.Morph(c), typedef.New.Morph(c))
+	c.Putln("type %s %s", typedef.New.Morph(c), typedef.Old.Morph(c))
 }
 
 // Struct morphing.
 func (strct *Struct) Morph(c *Context) {
+	c.Putln("type %s struct {", strct.Name.Morph(c))
+	strct.Fields.MorphDefine(c)
+	c.Putln("}")
+	c.Putln("\n")
 }
 
 // Union morphing.
 func (union *Union) Morph(c *Context) {
+	c.Putln("type %s struct {", union.Name.Morph(c))
+	union.Fields.MorphDefine(c)
+	c.Putln("}")
+	c.Putln("\n")
 }
 
 // Request morphing.
@@ -150,10 +367,54 @@ func (request *Request) Morph(c *Context) {
 
 // Event morphing.
 func (ev *Event) Morph(c *Context) {
+	name := ev.Name.Morph(c)
+
+	c.Putln("const %s = %d", name, ev.Number)
+	c.Putln("")
+	c.Putln("type %sEvent struct {", name)
+	ev.Fields.MorphDefine(c)
+	c.Putln("}")
+	c.Putln("")
+	c.Putln("func New%s(buf []byte) %sEvent {", name, name)
+	c.Putln("var v %sEvent", name)
+	ev.Fields.MorphRead(c, FieldsEvent, ev.NoSequence)
+	c.Putln("return v")
+	c.Putln("}")
+	c.Putln("")
+	c.Putln("func (err %sEvent) ImplementsEvent() { }", name)
+	c.Putln("")
+	c.Putln("func (ev %sEvent) Bytes() []byte {", name)
+	// ev.Fields.MorphWrite(c, FieldsEvent) 
+	c.Putln("}")
+	c.Putln("")
+	c.Putln("func init() {")
+	c.Putln("newEventFuncs[%d] = New%s", ev.Number, name)
+	c.Putln("}")
+	c.Putln("")
 }
 
 // EventCopy morphing.
 func (evcopy *EventCopy) Morph(c *Context) {
+	oldName, newName := evcopy.Ref.Morph(c), evcopy.Name.Morph(c)
+
+	c.Putln("const %s = %d", newName, evcopy.Number)
+	c.Putln("")
+	c.Putln("type %sEvent %sEvent", newName, oldName)
+	c.Putln("")
+	c.Putln("func New%s(buf []byte) %sEvent {", newName, newName)
+	c.Putln("return (%sEvent)(New%s(buf))", newName, oldName)
+	c.Putln("}")
+	c.Putln("")
+	c.Putln("func (err %sEvent) ImplementsEvent() { }", newName)
+	c.Putln("")
+	c.Putln("func (ev %sEvent) Bytes() []byte {", newName)
+	c.Putln("return (%sEvent)(ev).Bytes()", oldName)
+	c.Putln("}")
+	c.Putln("")
+	c.Putln("func init() {")
+	c.Putln("newEventFuncs[%d] = New%s", evcopy.Number, newName)
+	c.Putln("}")
+	c.Putln("")
 }
 
 // Error morphing.
@@ -162,6 +423,26 @@ func (err *Error) Morph(c *Context) {
 
 // ErrorCopy morphing.
 func (errcopy *ErrorCopy) Morph(c *Context) {
+	oldName, newName := errcopy.Ref.Morph(c), errcopy.Name.Morph(c)
+
+	c.Putln("const Bad%s = %d", newName, errcopy.Number)
+	c.Putln("")
+	c.Putln("type %sError %sError", newName, oldName)
+	c.Putln("")
+	c.Putln("func New%sError(buf []byte) %sError {", newName, newName)
+	c.Putln("return (%sError)(New%sError(buf))", newName, oldName)
+	c.Putln("}")
+	c.Putln("")
+	c.Putln("func (err %sError) ImplementsError() { }", newName)
+	c.Putln("")
+	c.Putln("func (err %sError) Bytes() []byte {", newName)
+	c.Putln("return (%sError)(err).Bytes()", oldName)
+	c.Putln("}")
+	c.Putln("")
+	c.Putln("func init() {")
+	c.Putln("newErrorFuncs[%d] = New%sError", errcopy.Number, newName)
+	c.Putln("}")
+	c.Putln("")
 }
 
 /******************************************************************************/
diff --git a/nexgb/xgbgen/main.go b/nexgb/xgbgen/main.go
index 69579a4..c69c8aa 100644
--- a/nexgb/xgbgen/main.go
+++ b/nexgb/xgbgen/main.go
@@ -55,6 +55,7 @@ func main() {
 		cmdGofmt := exec.Command("gofmt")
 		cmdGofmt.Stdin = c.out
 		cmdGofmt.Stdout = os.Stdout
+		cmdGofmt.Stderr = os.Stderr
 		err = cmdGofmt.Run()
 		if err != nil {
 			log.Fatal(err)
diff --git a/nexgb/xgbgen/misc.go b/nexgb/xgbgen/misc.go
index 9adcf5d..13c4cc2 100644
--- a/nexgb/xgbgen/misc.go
+++ b/nexgb/xgbgen/misc.go
@@ -21,6 +21,11 @@ func popCount(mask uint) uint {
 	return n
 }
 
+// pad makes sure 'n' aligns on 4 bytes.
+func pad(n int) int {
+	return (n + 3) & ^3
+}
+
 // splitAndTitle takes a string, splits it by underscores, capitalizes the
 // first letter of each chunk, and smushes'em back together.
 func splitAndTitle(s string) string {
@@ -29,9 +34,9 @@ func splitAndTitle(s string) string {
 		return strings.Title(strings.ToLower(s))
 	}
 
-	// If the string has no underscores, leave it be.
+	// If the string has no underscores, capitalize it and leave it be.
 	if i := strings.Index(s, "_"); i == -1 {
-		return s
+		return strings.Title(s)
 	}
 
 	// Now split the name at underscores, capitalize the first
diff --git a/nexgb/xgbgen/xgbgen b/nexgb/xgbgen/xgbgen
deleted file mode 100755
index ef33abc..0000000
--- a/nexgb/xgbgen/xgbgen
+++ /dev/null
diff --git a/nexgb/xgbgen/xml.go b/nexgb/xgbgen/xml.go
index 0f632b4..12932b2 100644
--- a/nexgb/xgbgen/xml.go
+++ b/nexgb/xgbgen/xml.go
@@ -71,16 +71,16 @@ func (x *XML) Morph(c *Context) {
 	x.Requests.Morph(c)
 	c.Putln("")
 
-	x.Events.Morph(c)
+	x.Errors.Morph(c)
 	c.Putln("")
 
-	x.Errors.Morph(c)
+	x.ErrorCopies.Morph(c)
 	c.Putln("")
 
-	x.EventCopies.Morph(c)
+	x.Events.Morph(c)
 	c.Putln("")
 
-	x.ErrorCopies.Morph(c)
+	x.EventCopies.Morph(c)
 	c.Putln("")
 }
 
@@ -166,6 +166,38 @@ type Name string
 
 type Type string
 
+// Size is a nifty function that takes any type and digs until it finds
+// its underlying base type. At which point, the size can be determined.
+func (typ Type) Size(c *Context) uint {
+	// If this is a base type, we're done.
+	if size, ok := BaseTypeSizes[string(typ)]; ok {
+		return size
+	}
+
+	// If it's a resource, we're also done.
+	if c.xml.IsResource(typ) {
+		return BaseTypeSizes["Id"]
+	}
+
+	// It's not, so that implies there is *some* typedef declaring it
+	// in terms of another type. Just follow that chain until we get to
+	// a base type. We also need to check imported stuff.
+	for _, typedef := range c.xml.TypeDefs {
+		if typ == typedef.New {
+			return typedef.Old.Size(c)
+		}
+	}
+	for _, imp := range c.xml.Imports {
+		for _, typedef := range imp.xml.TypeDefs {
+			if typ == typedef.New {
+				return typedef.Old.Size(c)
+			}
+		}
+	}
+	log.Panicf("Could not find base size of type '%s'.", typ)
+	panic("unreachable")
+}
+
 type Imports []*Import
 
 func (imports Imports) Eval() {
@@ -239,7 +271,7 @@ type EventCopies []*EventCopy
 
 type EventCopy struct {
 	Name Type `xml:"name,attr"`
-	Number string `xml:"number,attr"`
+	Number int `xml:"number,attr"`
 	Ref Type `xml:"ref,attr"`
 }
 
@@ -247,7 +279,7 @@ type ErrorCopies []*ErrorCopy
 
 type ErrorCopy struct {
 	Name Type `xml:"name,attr"`
-	Number string `xml:"number,attr"`
+	Number int `xml:"number,attr"`
 	Ref Type `xml:"ref,attr"`
 }
 
@@ -255,14 +287,14 @@ type Structs []*Struct
 
 type Struct struct {
 	Name Type `xml:"name,attr"`
-	Fields []*Field `xml:",any"`
+	Fields Fields `xml:",any"`
 }
 
 type Unions []*Union
 
 type Union struct {
 	Name Type `xml:"name,attr"`
-	Fields []*Field `xml:",any"`
+	Fields Fields `xml:",any"`
 }
 
 type Requests []*Request
@@ -271,12 +303,12 @@ type Request struct {
 	Name Type `xml:"name,attr"`
 	Opcode int `xml:"opcode,attr"`
 	Combine bool `xml:"combine-adjacent,attr"`
-	Fields []*Field `xml:",any"`
+	Fields Fields `xml:",any"`
 	Reply *Reply `xml:"reply"`
 }
 
 type Reply struct {
-	Fields []*Field `xml:",any"`
+	Fields Fields `xml:",any"`
 }
 
 type Events []*Event
@@ -285,7 +317,7 @@ type Event struct {
 	Name Type `xml:"name,attr"`
 	Number int `xml:"number,attr"`
 	NoSequence bool `xml:"no-sequence-number,true"`
-	Fields []*Field `xml:",any"`
+	Fields Fields `xml:",any"`
 }
 
 type Errors []*Error
@@ -293,6 +325,6 @@ type Errors []*Error
 type Error struct {
 	Name Type `xml:"name,attr"`
 	Number int `xml:"number,attr"`
-	Fields []*Field `xml:",any"`
+	Fields Fields `xml:",any"`
 }
 
diff --git a/nexgb/xgbgen/xml_fields.go b/nexgb/xgbgen/xml_fields.go
index 18be6e3..c1a7240 100644
--- a/nexgb/xgbgen/xml_fields.go
+++ b/nexgb/xgbgen/xml_fields.go
@@ -23,6 +23,8 @@ import (
 	"strings"
 )
 
+type Fields []*Field
+
 type Field struct {
 	XMLName xml.Name
 
@@ -30,7 +32,7 @@ type Field struct {
 	Bytes int `xml:"bytes,attr"`
 
 	// For 'field', 'list', 'localfield', 'exprfield' and 'switch' elements.
-	Name string `xml:"name,attr"`
+	Name Name `xml:"name,attr"`
 
 	// For 'field', 'list', 'localfield', and 'exprfield' elements.
 	Type Type `xml:"type,attr"`
@@ -40,11 +42,11 @@ type Field struct {
 
 	// For 'valueparm' element.
 	ValueMaskType Type `xml:"value-mask-type,attr"`
-	ValueMaskName string `xml:"value-mask-name,attr"`
-	ValueListName string `xml:"value-list-name,attr"`
+	ValueMaskName Name `xml:"value-mask-name,attr"`
+	ValueListName Name `xml:"value-list-name,attr"`
 
 	// For 'switch' element.
-	Bitcases []*Bitcase `xml:"bitcase"`
+	Bitcases Bitcases `xml:"bitcase"`
 
 	// I don't know which elements these are for. The documentation is vague.
 	// They also seem to be completely optional.
@@ -86,6 +88,8 @@ func (f *Field) String() string {
 	panic("unreachable")
 }
 
+type Bitcases []*Bitcase
+
 // Bitcase represents a single expression followed by any number of fields.
 // Namely, if the switch's expression (all bitcases are inside a switch),
 // and'd with the bitcase's expression is equal to the bitcase expression,
@@ -95,7 +99,7 @@ func (f *Field) String() string {
 // it's the closest thing to a Union I can get to in Go without interfaces.
 // Would an '<expression>' tag have been too much to ask? :-(
 type Bitcase struct {
-	Fields []*Field `xml:",any"`
+	Fields Fields `xml:",any"`
 
 	// All the different expressions.
 	// When it comes time to choose one, use the 'Expr' method.