Vulkan-Docs/src/spec/registry.rnc

# Copyright (c) 2013-2018 The Khronos Group Inc.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
#     http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.

# Relax NG schema for Khronos Vulkan API Registry XML
#
# See https://www.khronos.org/vulkan/
#
# This definition is subject to change (mostly in the form of additions)

namespace xsd = "http://www.w3.org/2001/XMLSchema-datatypes"

# Toplevel is a <registry> tag.
# May be led by an optional <comment> tag containing e.g. copyrights.
start = element registry {
    (
        element comment { text } ? |
        Vendorids  * |
        Tags       * |
        Types      * |
        Enums      * |
        Commands   * |
        Feature    * |
        Extensions *
    ) *
}

# <vendorids> defines a group of vendor IDs
Vendorids = element vendorids {
    Comment ? ,
    Vendorid *
}

# <vendorid> defines a single vendor ID.
#   name - author ID of the vendor
#   id - Khronos vendor ID (hexadecimal constant starting at 0x10000)
#   comment - unused
Vendorid = element vendorid {
    attribute name { text } ,
    attribute id { text } ,
    Comment ?
}

# <tags> defines a group of author tags
Tags = element tags {
    Comment ? ,
    Tag *
}

# <tag> defines a single author tag.
#   name - name of the tag
#   author - name of the author (usually a company or project name)
#   contact - contact responsible for the tag (name and contact information)
Tag = element tag {
    attribute name { text } ,
    attribute author { text } ,
    attribute contact { text }
}

# <types> defines a group of types
Types = element types {
    Comment ? ,
    (
        Type |
        element comment { text }
    ) *
}

# <type> defines a single type. It is usually a C typedef but
# may contain arbitrary C code.
#   name - name of this type, if not present in the <name> tag
#   api - matches a <feature> api attribute, if present
#   requires - name of another type definition required by this one
#   category - if present, 'enum' indicates a matching <enums>
#       block to generate an enumerated type for, and 'struct'
#       causes special interpretation of the contents of the type
#       tag including ... TBD ...
#       Other allowed values are 'include', 'define', 'handle' and 'bitmask',
#       which don't change syntactic interpretation but allow organization in
#       the generated header.
#   comment - unused
#   parent - only applicable if category is 'handle'. Notes another type with
#       the 'handle' category that acts as a parent object for this type.
#   returnedonly - only applicable if category is 'struct'. Notes that this
#       struct is going to be filled in by the API, rather than an application
#       filling it out and passing it to the API.
#   structextends - only applicable if category is 'struct'. Lists parent
#       structures which this structure may extend via the pNext chain
#       of the parent.
# For types without a category, contents include
#   <apientry /> - substitutes for an APIENTRY-style macro on output
#   <name> - contains name of the type being defined
#   <type> - contains name of types used to define this type. There
#       may be multiple imbedded <type> tags
# For types with category 'enum', contents should be empty
# For types with category 'struct', contents should be one or more
#   <member> - like <param> for a struct or union member
#       len - if the member is an array, len may be one or more of the following
#           things, separated by commas (one for each array indirection): another
#           member of that struct, 'null-terminated' for a string, '1' to indicate it's
#           just a pointer (used for nested pointers), or a latex equation (prefixed with
#           'latexmath:')
#       altlen - if len has latexmath equations, this contains equivalent C99
#                expressions separated by commas.
#       externsync - denotes that the member should be externally synchronized
#           when accessed by Vulkan
#       optional - whether this value can be omitted by providing NULL (for
#           pointers), VK_NULL_HANDLE (for handles) or 0 (for bitmasks/values)
#       noautovalidity - tag stating that no automatic validity language should be generated
#       values - comma-separated list of legal values, usually used only for sType enums
#   <comment> - containing arbitrary text (unused)
#
# *** There's a problem here: I'm not sure how to represent the <type>
# syntax where it may contain arbitrarily interleaved text, <type>, and
# <enum> child tags. This allows only the syntax
#   text <type>name</type> text <enum>name</enum> text
# where <type> and <enum> are both optional and occur in the specified
# order, which might eventually be a problem.
Type = element type {
    attribute api { text } ? ,
    attribute requires { text } ? ,
    attribute name { TypeName } ? ,
    attribute category { text } ? ,
    attribute parent { TypeName } ? ,
    attribute returnedonly { text } ? ,
    attribute structextends { text } ? ,
    Comment ? ,
    (
        (
            ( text ,
              element type { text } *
            ) * ,
            element apientry { text } ? ,
            ( text ,
              element type { text } *
            ) * ,
            element name { TypeName } ? ,
            ( text ,
              element type { text } *
            ) *
        ) |
        (
            element member {
                attribute len { text } ? ,
                attribute altlen { text } ? ,
                attribute externsync { text } ? ,
                attribute optional { text } ? ,
                attribute noautovalidity { text } ? ,
                attribute values { text } ? ,
                mixed {
                    element type { TypeName } ? ,
                    element name { text } ,
                    element enum { EnumName } ? ,
                    element comment { text } ?
                }
            } |
            element comment { text }
        ) *
    )
}

# <enums> defines a group of enumerants
#   name - identifies a type name associated with this group. Should
#       match a <type> name to trigger generation of the type.
#   start, end - beginning and end of a numeric range
#   vendor - owner of the numeric range
#   type - 'enum' or 'bitmask', if present
#   comment - unused
Enums = element enums {
    attribute name { text } ? ,
    attribute type { text } ? ,
    attribute start { Integer } ? ,
    attribute end { Integer } ? ,
    Vendor ? ,
    Comment ? ,
    (
        Enum |
        Unused |
        element comment { text}
    ) *
}

# <enum> defines or references a single enumerant. There are two places it
# can be used: in an <enums> block, providing a global definition which
# may later be required by a feature or extension; or in a feature or
# extension, defining an enumerant specific to that feature. The second
# form has more possible attributes. Some combinations of attributes are
# nonsensical in on or the other place, but these are not detected by the
# validator.
#
# Ways to specify the enumerant value:
#   value - integer (including hex) value of the enumerant
#   bitpos - integer bit position of the enumerant in a bitmask
#   offset, [dir] - integer offset and direction ('-' for negative,
#       or positive if not specified) from a base value
#
# value and bitpos allow, and offset/dir require, the attribute:
#   extends - type name of the enumerant being extended
#
# Other attributes:
#   api - matches a <feature> api attribute, if present
#   type - 'u' (unsigned), 'ull' (uint64), or integer if not present
#   name - enumerant name
#   alias - another enumerant this is semantically identical to
#   comment - unused
Enum = element enum {
    (
      (
        (
          attribute value { Integer } &
          attribute extends { TypeName } ?
        ) |
        (
          attribute bitpos { Integer } &
          attribute extends { TypeName } ?
        ) |
        (
          attribute offset { Integer } &
          attribute dir { text } ? &
          attribute extends { TypeName }
        )
      ) ? &
      attribute api { text } ? &
      attribute type { TypeSuffix } ? &
      attribute name { text } &
      attribute alias { text } ? &
      Comment ?
    )
}

# <unused> defines a range of enumerants not currently being used
#   start, end - beginning and end of an unused numeric range
#   vendor - unused
#   comment - unused
Unused = element unused {
    attribute start { Integer } ,
    attribute end { Integer } ? ,
    Vendor ? ,
    Comment ?
}
# <commands> defines a group of commands
Commands = element commands {
    Comment ? ,
    Command *
}

# <command> defines a single command
#
# The possible attributes are not described in this comment block yet, but
# are in readme.pdf.
#
#   <proto> is the C function prototype, including the return type
#   <param> are function parameters, in order
#     len - if the member is an array, len may be one or more of the following
#           things, separated by commas (one for each array indirection): another
#           member of that struct, 'null-terminated' for a string, '1' to indicate it's
#           just a pointer (used for nested pointers), or a latex equation (prefixed with
#           'latexmath:')
#     altlen - if len has latexmath equations, this contains equivalent C99
#              expressions separated by commas.
#     externsync - denotes that the member should be externally synchronized
#         when accessed by Vulkan
#     optional - whether this value can be omitted by providing NULL (for
#         pointers), VK_NULL_HANDLE (for handles) or 0 (for bitmasks/values)
#     noautovalidity - tag stating that no automatic validity language should be generated
#     <type> is a <type> name, if present
#     <name> is the function / parameter name
# The textual contents of <proto> and <param> should be legal C
# for those parts of a function declaration.
#   <alias> - denotes function aliasing, if present
#     name - name of aliased function
#   <description> - unused text
#   <implicitexternsyncparams> are spec-language descriptions of
#       objects that are not parameters of the command, but
#       are related to them and also require external synchronization.
Command = element command {
    attribute queues { text } ? ,
    attribute successcodes { text } ? ,
    attribute errorcodes { text } ? ,
    attribute renderpass { text } ? ,
    attribute cmdbufferlevel { text } ? ,
    attribute pipeline { text } ? ,
    Comment ? ,
    element proto {
        mixed {
            element type { TypeName } ? ,
            element name { text }
        }
    } ,
    element param {
        attribute len { text } ? ,
        attribute altlen { text } ? ,
        attribute externsync { text } ? ,
        attribute optional { text } ? ,
        attribute noautovalidity { text } ? ,
        mixed {
            element type { TypeName } ? ,
            element name { text }
        }
    } * ,
    (
        element alias {
            Name
        } ? &
        element description {
            text
        } ? &
        element implicitexternsyncparams {
            element param { text } *
        } ?
    )
}

# Each <feature> defines the interface of an API version (e.g. OpenGL 1.2)
#   api - API tag (e.g. 'gl', 'gles2', etc. - used internally, not
#     necessarily an actual API name
#   name - version name (C preprocessor name, e.g. GL_VERSION_4_2)
#   number - version number, e.g. 4.2
#   protect - additional #ifdef symbol to place around the feature
#   <require> / <remove> contains features to require or remove in
#                        this version
#     profile - only require/remove when generated profile matches
#     comment - unused
Feature = element feature {
    attribute api { text } ,
    Name ,
    attribute number { xsd:float } ,
    attribute protect { text } ? ,
    Comment ? ,
    (
        element require {
            ProfileName ? ,
            ExtensionName ? ,
            Comment ? ,
            (
                InterfaceElement |
                element comment { text }
            ) *
        } |
        element remove {
            ProfileName ? ,
            Comment ? ,
            (
                InterfaceElement |
                element comment { text }
            ) *
        }
    ) *
}

Extensions = element extensions {
    Comment ? ,
    Extension *
}

# Defines the interface of an API <extension>. Like a <feature>
# tag, but with slightly different attributes:
#   api - regexp pattern matching one or more API tags, indicating
#     which APIs the extension is known to work with. The only
#     syntax supported is <name>{|<name>}* and each name must
#     exactly match an API being generated (implicit ^$ surrounding).
#   name - extension name string
#   number - extension number (positive integer, should be unique)
#   protect - C preprocessor symbol to conditionally define the interface
#   author - name of the author (usually a company or project name)
#   contact - contact responsible for the tag (name and contact information)
#   type - 'device' or 'instance', if present
#   requires - commas-separated list of extension names required by this
#       extension
#   supported - profile name(s) supporting this extension, e.g. 'vulkan'
#       or 'disabled' to never generate output.
# In addition, <require> / <remove> tags also support an api attribute:
#     api - only require/remove these features for the matching API.
#       Not a regular expression.
Extension = element extension {
    Name ,
    attribute number { Integer } ? ,
    attribute protect { text } ? ,
    attribute author { text } ? ,
    attribute contact { text } ? ,
    attribute type { text } ? ,
    attribute requires { text } ? ,
    attribute supported { StringGroup } ? ,
    Comment ? ,
    (
        element require {
            attribute api { text } ? ,
            ProfileName ? ,
            ExtensionName ? ,
            Comment ? ,
            (
                InterfaceElement |
                element comment { text }
            ) *
        } |
        element remove {
            attribute api { text } ? ,
            ProfileName ? ,
            Comment ? ,
            (
                InterfaceElement |
                element comment { text }
            ) *
        }
    ) *
}

# Contents of a <require> / <remove> tag, defining a group
# of features to require or remove.
#   <type> / <enum> / <command> all have attributes
#     name - feature name which must match
InterfaceElement =
    element type {
        Name ,
        Comment ?
    } |
    Enum |
    element command {
        Name ,
        Comment ?
    }

# Integers are allowed to be either decimal or C-hex (0x[0-9A-F]+), but
# XML Schema types don't seem to support hex notation, so we use this
# as a placeholder.
Integer = text

# EnumName is an compile-time constant name
EnumName = text

# TypeName is an argument/return value C type name
TypeName = text

# TypeSuffix is a C numeric type suffix, e.g. 'u' or 'ull'
TypeSuffix = text

# StringGroup is a regular expression with an implicit
#   '^(' and ')$' bracketing it.
StringGroup = text

# Repeatedly used attributes
ProfileName = attribute profile { text }
ExtensionName = attribute extension { text }
Vendor = attribute vendor { text }
Comment = attribute comment { text }
Name = attribute name { text }