diff options
Diffstat (limited to 'files/zh-cn/mozilla/tech/xpidl/syntax/index.html')
| -rw-r--r-- | files/zh-cn/mozilla/tech/xpidl/syntax/index.html | 127 |
1 files changed, 127 insertions, 0 deletions
diff --git a/files/zh-cn/mozilla/tech/xpidl/syntax/index.html b/files/zh-cn/mozilla/tech/xpidl/syntax/index.html new file mode 100644 index 0000000000..daca23f49f --- /dev/null +++ b/files/zh-cn/mozilla/tech/xpidl/syntax/index.html @@ -0,0 +1,127 @@ +--- +title: Syntax +slug: Mozilla/Tech/XPIDL/Syntax +tags: + - xpidl +translation_of: Mozilla/Tech/XPIDL/Syntax +--- +<h3 id="Status_of_this_document" name="Status_of_this_document">Status of this document</h3> + +<p>This is a partial reverse-engineering of the libIDL source code's parser, limited mostly to the subset of functionality that is supported by the Mozilla xpidl binary.</p> + +<h3 id="Purpose_of_this_document" name="Purpose_of_this_document">Purpose of this document</h3> + +<p>This document is <strong>not</strong> an introduction to XPIDL or IDL in general. It is more focused on XPIDL syntax and grammar. See <a href="/en/XPIDL" title="en/XPIDL">XPIDL Main Page</a> for more links and introductory content.</p> + +<h3 id="Simplifications.2C_conventions_and_notation" name="Simplifications.2C_conventions_and_notation">Simplifications, conventions and notation</h3> + +<p>The syntax is specified according to ABNF as defined by <a class="external" href="http://tools.ietf.org/html/rfc5234" title="http://tools.ietf.org/html/rfc5234">RFC 5234</a>, although a few productions use prose for clarity of understanding.</p> + +<p>Lexically, tokens are delimited by whitespace (defined here as spaces, tabs, vertical tabs, form feeds, line feeds, and carriage returns, or [ \t\v\f\r\n] in regular expression form). LibIDL only considers a single line feed as a newline, and not carriage returns (although xpidl begs to differ). Additionally, the use of both C-style (/* ... */) and C++-style (// ... end-of-line) comments are permitted between any two tokens.</p> + +<p>Some productions can only occur at the beginning of lines; to simplify the grammar, I will not mention them in the grammar, especially since they are handled as a preprocessing step before the IDL source code is actually parsed.</p> + +<ol> + <li>A `<code>%{</code>' that appears at the beginning of a line is the start of a <strong>raw code fragment</strong>, which extends until the end of a line that begins with `<code>%}</code>'. Text inside raw code fragments are not otherwise parsed by xpidl directly. This may be followed by the language, as in `<code>%{C++</code>', to output the raw fragment only in the specified language.</li> + <li>A `<code>#include "file"</code>' line instructs the xpidl processor to include that file in the same sense that the C preprocessor includes a file. Note that includes within comments or raw code fragments are not processed by xpidl. Unlike the C preprocessor, when a file is included multiple times, it acts as if the subsequent includes did not happen; this prevents the need for include guards.</li> +</ol> + +<h3 id="XPIDL_Syntax_.28BNF.29" name="XPIDL_Syntax_.28BNF.29">XPIDL Syntax (ABNF)</h3> + +<p>The root production here is <code>idl_file</code>.</p> + +<p><code><code>idl_file = 1*definition<br> + definition = [type_decl / const_decl / interface] ";"<br> + interface = [prop_list] "interface" ident [[inheritance] "{" *(ifacebody) "}"]<br> + inheritance = ":" *(scoped_name ",") scoped_name]<br> + ifacebody = [type_decl / op_decl /attr_decl / const_decl] ";" / codefrag<br> + <br> + type_decl = [prop_list] "typedef" type_spec *(ident ",") ident<br> + type_decl /= [prop_list] "native" ident [parens]<br> + const_decl = "const" type_spec ident "=" expr<br> + op_decl = [prop_list] (type_spec / "void") parameter_decls raise_list<br> + parameter_decls = "(" [*(param_decl ",") param_decl] ")"<br> + param_decl = [prop_list] ("in" / "out" / "inout") type_spec ident<br> + attr_decl = [prop_list] ["readonly"] "attribute" type_spec *(ident ",") ident<br> + <br> + ; Descending order of precedence<br> + expr /= expr ("|" / "^" / "&") expr ; Unequal precedence "|" is lowest<br> + expr /= expr ("<<" / ">>") expr<br> + expr /= expr ("+" / "-") expr<br> + expr /= expr ("*" / "/" / "%") expr<br> + expr /= ["-" / "+" / "~"] (scoped_name / literal / "(" expr ")" )<br> + <br> + ; Numeric literals: quite frankly, I'm sure you know how these kinds of<br> + ; literals work, and these are annoying to specify in ABNF.<br> + literal = octal_literal / decimal_literal / hex_literal / floating_literal<br> + literal /= string_literal / char_literal<br> + literal /= "TRUE" / "FALSE"<br> + <br> + ; In regex: /"[^"\n]*["\n]/. Yes, newline terminates.<br> + string_literal = 1*(%x22 *(any char except %x22 or %x0a) (%x22 / %x0a))<br> + ; Same as above, but s/"/'/g<br> + char_literal = 1*(%x27 *(any char except %x27 or %x0a) (%x27 / %x0a))<br> + <br> + type_spec = "float" / "double" / "string" / "wstring"<br> + type_spec /= ["unsigned"] ("short" / "long" / "long" "long")<br> + type_spec /= "char" / "wchar" / "boolean" / "octet"<br> + type_spec /= scoped_name<br> + <br> + prop_list = "[" *(property ",") property "]"<br> + property = ident [parens]<br> + raise_list = "raises" "(" *(scoped_name) ",") scoped_name ")"<br> + <br> + scoped_name = *(ident "::") ident / "::" ident</code></code><br> + <code><code>; In regex: [A-Za-z_][A-Za-z0-9_]*; identifiers beginning with _ cause warnings</code></code><br> + <code><code>ident = (%x41-5a / %x61-7a / "_") *(%x41-5a / %x61-7a / %x30-39 / "_")<br> + parens = "(" 1*(any char except ")") ")"</code></code></p> + +<h3 id="Functionality_not_used_in_xpidl">Functionality not used in xpidl</h3> + +<p>The libIDL parser we use is more powerful than xpidl itself can understand. The following is a list of potential features which are parseable but may not result in expected code:</p> + +<ul> + <li>Struct, union, and enumerated types</li> + <li>Array declarators (appears to be supported in xpidl_header.c but not xpidl_typelib.c)</li> + <li>Exception declarations</li> + <li>Module declarations</li> + <li>Variable arguments (that makes the ABNF get more wonky)</li> + <li>Sequence types</li> + <li>Max-length strings</li> + <li>Fixed-point numbers</li> + <li>"any" and "long double" types.</li> +</ul> + +<h3 id="Pyxpidl_syntax">Pyxpidl syntax</h3> + +<p><code><code>idlfile = *(CDATA / INCLUDE / interface / typedef / native)<br> + <br> + typedef = "typedef" IDENTIFER IDENTIFIER ";"<br> + native = [attributes] "native" IDENTIFIER "(" NATIVEID ")"<br> + interface = [attributes] "interface" IDENTIFIER" [ifacebase] [ifacebody] ";"<br> + ifacebase = ":" IDENTIFIER<br> + ifacebody = "{" *(member) "}"<br> + <br> + member = CDATA / "const" IDENTIFIER IDENTIFIER "=" number ";"<br> + member /= [attributes] ["readonly"] "attribute" IDENTIFIER IDENTIFER ";"<br> + member /= [attributes] IDENTIFIER IDENTIFIER "(" paramlist ")" raises ";"<br> + paramlist = [param *("," param)]<br> + raises = ["raises" "(" IDENTIFIER *("," identifier) ")"]<br> + attributes = "[" attribute *("," attribute) "]"<br> + attribute = (IDENTIFIER / CONST) ["(" (IDENTIFIER / IID) ")"]<br> + param = [attributes] ("in" / "out" / "inout") IDENTIFIER IDENTIFIER<br> + <br> + number = NUMBER / IDENTIFIER<br> + number /= "(" number ")"<br> + number /= "-" number<br> + number /= number ("+" / "-" / "*") number<br> + number /= number ("<<" / >>") number<br> + number /= number "|" number<br> + <br> + ; Lexical tokens, I'm going to specify these in regex form<br> + NUMBER = /-?\d+|0x[0-9A-Fa-f]+/<br> + CDATA = /%\{[ ]*C\+\+[ ]*\n(.*?\n?)%\}[ ]*(C\+\+)?/s<br> + INCLUDE = /\#include[ \t]+"[^"\n]+"/<br> + NATIVEID = /[^()\n]+(?=\))/<br> + IID = /[0-9A-Fa-f]{8}-[0-9A-Fa-f]{4}-[0-9A-Fa-f]{4}-[0-9A-Fa-f]{12}/<br> + IDENTIFIER = /unsigned long long|unsigned short|unsigned long|long long|[A-Za-z][A-Za-z_0-9]*/</code></code></p> |