| 1 |
"""A simple non-validating parser for C99. |
| 2 |
|
| 3 |
The functions and regex patterns here are not entirely suitable for |
| 4 |
validating C syntax. Please rely on a proper compiler for that. |
| 5 |
Instead our goal here is merely matching and extracting information from |
| 6 |
valid C code. |
| 7 |
|
| 8 |
Furthermore, the grammar rules for the C syntax (particularly as |
| 9 |
described in the K&R book) actually describe a superset, of which the |
| 10 |
full C language is a proper subset. Here are some of the extra |
| 11 |
conditions that must be applied when parsing C code: |
| 12 |
|
| 13 |
* ... |
| 14 |
|
| 15 |
(see: https://www.open-std.org/jtc1/sc22/wg14/www/docs/n1256.pdf) |
| 16 |
|
| 17 |
We have taken advantage of the elements of the C grammar that are used |
| 18 |
only in a few limited contexts, mostly as delimiters. They allow us to |
| 19 |
focus the regex patterns confidently. Here are the relevant tokens and |
| 20 |
in which grammar rules they are used: |
| 21 |
|
| 22 |
separators: |
| 23 |
* ";" |
| 24 |
+ (decl) struct/union: at end of each member decl |
| 25 |
+ (decl) declaration: at end of each (non-compound) decl |
| 26 |
+ (stmt) expr stmt: at end of each stmt |
| 27 |
+ (stmt) for: between exprs in "header" |
| 28 |
+ (stmt) goto: at end |
| 29 |
+ (stmt) continue: at end |
| 30 |
+ (stmt) break: at end |
| 31 |
+ (stmt) return: at end |
| 32 |
* "," |
| 33 |
+ (decl) struct/union: between member declators |
| 34 |
+ (decl) param-list: between params |
| 35 |
+ (decl) enum: between enumerators |
| 36 |
+ (decl) initializer (compound): between initializers |
| 37 |
+ (expr) postfix: between func call args |
| 38 |
+ (expr) expression: between "assignment" exprs |
| 39 |
* ":" |
| 40 |
+ (decl) struct/union: in member declators |
| 41 |
+ (stmt) label: between label and stmt |
| 42 |
+ (stmt) case: between expression and stmt |
| 43 |
+ (stmt) default: between "default" and stmt |
| 44 |
* "=" |
| 45 |
+ (decl) declaration: between decl and initializer |
| 46 |
+ (decl) enumerator: between identifier and "initializer" |
| 47 |
+ (expr) assignment: between "var" and expr |
| 48 |
|
| 49 |
wrappers: |
| 50 |
* "(...)" |
| 51 |
+ (decl) declarator (func ptr): to wrap ptr/name |
| 52 |
+ (decl) declarator (func ptr): around params |
| 53 |
+ (decl) declarator: around sub-declarator (for readability) |
| 54 |
+ (expr) postfix (func call): around args |
| 55 |
+ (expr) primary: around sub-expr |
| 56 |
+ (stmt) if: around condition |
| 57 |
+ (stmt) switch: around source expr |
| 58 |
+ (stmt) while: around condition |
| 59 |
+ (stmt) do-while: around condition |
| 60 |
+ (stmt) for: around "header" |
| 61 |
* "{...}" |
| 62 |
+ (decl) enum: around enumerators |
| 63 |
+ (decl) func: around body |
| 64 |
+ (stmt) compound: around stmts |
| 65 |
* "[...]" |
| 66 |
* (decl) declarator: for arrays |
| 67 |
* (expr) postfix: array access |
| 68 |
|
| 69 |
other: |
| 70 |
* "*" |
| 71 |
+ (decl) declarator: for pointer types |
| 72 |
+ (expr) unary: for pointer deref |
| 73 |
|
| 74 |
|
| 75 |
To simplify the regular expressions used here, we've takens some |
| 76 |
shortcuts and made certain assumptions about the code we are parsing. |
| 77 |
Some of these allow us to skip context-sensitive matching (e.g. braces) |
| 78 |
or otherwise still match arbitrary C code unambiguously. However, in |
| 79 |
some cases there are certain corner cases where the patterns are |
| 80 |
ambiguous relative to arbitrary C code. However, they are still |
| 81 |
unambiguous in the specific code we are parsing. |
| 82 |
|
| 83 |
Here are the cases where we've taken shortcuts or made assumptions: |
| 84 |
|
| 85 |
* there is no overlap syntactically between the local context (func |
| 86 |
bodies) and the global context (other than variable decls), so we |
| 87 |
do not need to worry about ambiguity due to the overlap: |
| 88 |
+ the global context has no expressions or statements |
| 89 |
+ the local context has no function definitions or type decls |
| 90 |
* no "inline" type declarations (struct, union, enum) in function |
| 91 |
parameters ~(including function pointers)~ |
| 92 |
* no "inline" type decls in function return types |
| 93 |
* no superfluous parentheses in declarators |
| 94 |
* var decls in for loops are always "simple" (e.g. no inline types) |
| 95 |
* only inline struct/union/enum decls may be anonymous (without a name) |
| 96 |
* no function pointers in function pointer parameters |
| 97 |
* for loop "headers" do not have curly braces (e.g. compound init) |
| 98 |
* syntactically, variable decls do not overlap with stmts/exprs, except |
| 99 |
in the following case: |
| 100 |
spam (*eggs) (...) |
| 101 |
This could be either a function pointer variable named "eggs" |
| 102 |
or a call to a function named "spam", which returns a function |
| 103 |
pointer that gets called. The only differentiator is the |
| 104 |
syntax used in the "..." part. It will be comma-separated |
| 105 |
parameters for the former and comma-separated expressions for |
| 106 |
the latter. Thus, if we expect such decls or calls then we must |
| 107 |
parse the decl params. |
| 108 |
""" |
| 109 |
|
| 110 |
""" |
| 111 |
TODO: |
| 112 |
* extract CPython-specific code |
| 113 |
* drop include injection (or only add when needed) |
| 114 |
* track position instead of slicing "text" |
| 115 |
* Parser class instead of the _iter_source() mess |
| 116 |
* alt impl using a state machine (& tokenizer or split on delimiters) |
| 117 |
""" |
| 118 |
|
| 119 |
from ..info import ParsedItem |
| 120 |
from ._info import SourceInfo |
| 121 |
|
| 122 |
|
| 123 |
def parse(srclines, **srckwargs): |
| 124 |
if isinstance(srclines, str): # a filename |
| 125 |
raise NotImplementedError |
| 126 |
|
| 127 |
anon_name = anonymous_names() |
| 128 |
for result in _parse(srclines, anon_name, **srckwargs): |
| 129 |
yield ParsedItem.from_raw(result) |
| 130 |
|
| 131 |
|
| 132 |
# XXX Later: Add a separate function to deal with preprocessor directives |
| 133 |
# parsed out of raw source. |
| 134 |
|
| 135 |
|
| 136 |
def anonymous_names(): |
| 137 |
counter = 1 |
| 138 |
def anon_name(prefix='anon-'): |
| 139 |
nonlocal counter |
| 140 |
name = f'{prefix}{counter}' |
| 141 |
counter += 1 |
| 142 |
return name |
| 143 |
return anon_name |
| 144 |
|
| 145 |
|
| 146 |
############################# |
| 147 |
# internal impl |
| 148 |
|
| 149 |
import logging |
| 150 |
|
| 151 |
|
| 152 |
_logger = logging.getLogger(__name__) |
| 153 |
|
| 154 |
|
| 155 |
def _parse(srclines, anon_name, **srckwargs): |
| 156 |
from ._global import parse_globals |
| 157 |
|
| 158 |
source = _iter_source(srclines, **srckwargs) |
| 159 |
for result in parse_globals(source, anon_name): |
| 160 |
# XXX Handle blocks here instead of in parse_globals(). |
| 161 |
yield result |
| 162 |
|
| 163 |
|
| 164 |
# We use defaults that cover most files. Files with bigger declarations |
| 165 |
# are covered elsewhere (MAX_SIZES in cpython/_parser.py). |
| 166 |
|
| 167 |
def _iter_source(lines, *, maxtext=11_000, maxlines=200, showtext=False): |
| 168 |
maxtext = maxtext if maxtext and maxtext > 0 else None |
| 169 |
maxlines = maxlines if maxlines and maxlines > 0 else None |
| 170 |
filestack = [] |
| 171 |
allinfo = {} |
| 172 |
# "lines" should be (fileinfo, data), as produced by the preprocessor code. |
| 173 |
for fileinfo, line in lines: |
| 174 |
if fileinfo.filename in filestack: |
| 175 |
while fileinfo.filename != filestack[-1]: |
| 176 |
filename = filestack.pop() |
| 177 |
del allinfo[filename] |
| 178 |
filename = fileinfo.filename |
| 179 |
srcinfo = allinfo[filename] |
| 180 |
else: |
| 181 |
filename = fileinfo.filename |
| 182 |
srcinfo = SourceInfo(filename) |
| 183 |
filestack.append(filename) |
| 184 |
allinfo[filename] = srcinfo |
| 185 |
|
| 186 |
_logger.debug(f'-> {line}') |
| 187 |
srcinfo._add_line(line, fileinfo.lno) |
| 188 |
if srcinfo.too_much(maxtext, maxlines): |
| 189 |
break |
| 190 |
while srcinfo._used(): |
| 191 |
yield srcinfo |
| 192 |
if showtext: |
| 193 |
_logger.debug(f'=> {srcinfo.text}') |
| 194 |
else: |
| 195 |
if not filestack: |
| 196 |
srcinfo = SourceInfo('???') |
| 197 |
else: |
| 198 |
filename = filestack[-1] |
| 199 |
srcinfo = allinfo[filename] |
| 200 |
while srcinfo._used(): |
| 201 |
yield srcinfo |
| 202 |
if showtext: |
| 203 |
_logger.debug(f'=> {srcinfo.text}') |
| 204 |
yield srcinfo |
| 205 |
if showtext: |
| 206 |
_logger.debug(f'=> {srcinfo.text}') |
| 207 |
if not srcinfo._ready: |
| 208 |
return |
| 209 |
# At this point either the file ended prematurely |
| 210 |
# or there's "too much" text. |
| 211 |
filename, lno, text = srcinfo.filename, srcinfo._start, srcinfo.text |
| 212 |
if len(text) > 500: |
| 213 |
text = text[:500] + '...' |
| 214 |
raise Exception(f'unmatched text ({filename} starting at line {lno}):\n{text}') |