OILS / mycpp / gc_mylib.cc View on Github | oilshell.org

297 lines, 160 significant
1#include "mycpp/gc_mylib.h"
2
3#include <errno.h>
4#include <stdio.h>
5#include <unistd.h> // isatty
6
7namespace mylib {
8
9void InitCppOnly() {
10 // We don't seem need this now that we have ctx_FlushStdout().
11 // setvbuf(stdout, 0, _IONBF, 0);
12
13 // Arbitrary threshold of 50K objects based on eyeballing
14 // benchmarks/osh-runtime 10K or 100K aren't too bad either.
15 gHeap.Init(50000);
16}
17
18void print_stderr(BigStr* s) {
19 fputs(s->data_, stderr); // prints until first NUL
20 fputc('\n', stderr);
21}
22
23#if 0
24void writeln(BigStr* s, int fd) {
25 // TODO: handle errors and write in a loop, like posix::write(). If possible,
26 // use posix::write directly, but that introduces some dependency problems.
27
28 if (write(fd, s->data_, len(s)) < 0) {
29 assert(0);
30 }
31 if (write(fd, "\n", 1) < 0) {
32 assert(0);
33 }
34}
35#endif
36
37BigStr* JoinBytes(List<int>* byte_list) {
38 int n = len(byte_list);
39 BigStr* result = NewStr(n);
40 for (int i = 0; i < n; ++i) {
41 result->data_[i] = byte_list->at(i);
42 }
43 return result;
44}
45
46// For BashArray
47void BigIntSort(List<mops::BigInt>* keys) {
48 keys->sort();
49}
50
51class MutableStr : public BigStr {};
52
53MutableStr* NewMutableStr(int n) {
54 // In order for everything to work, MutableStr must be identical in layout to
55 // BigStr. One easy way to achieve this is for MutableStr to have no members
56 // and to inherit from BigStr.
57 static_assert(sizeof(MutableStr) == sizeof(BigStr),
58 "BigStr and MutableStr must have same size");
59 return reinterpret_cast<MutableStr*>(NewStr(n));
60}
61
62Tuple2<BigStr*, BigStr*> split_once(BigStr* s, BigStr* delim) {
63 DCHECK(len(delim) == 1);
64
65 const char* start = s->data_; // note: this pointer may move
66 char c = delim->data_[0];
67 int length = len(s);
68
69 const char* p = static_cast<const char*>(memchr(start, c, length));
70
71 if (p) {
72 int len1 = p - start;
73 int len2 = length - len1 - 1; // -1 for delim
74
75 BigStr* s1 = nullptr;
76 BigStr* s2 = nullptr;
77 // Allocate together to avoid 's' moving in between
78 s1 = NewStr(len1);
79 s2 = NewStr(len2);
80
81 memcpy(s1->data_, s->data_, len1);
82 memcpy(s2->data_, s->data_ + len1 + 1, len2);
83
84 return Tuple2<BigStr*, BigStr*>(s1, s2);
85 } else {
86 return Tuple2<BigStr*, BigStr*>(s, nullptr);
87 }
88}
89
90LineReader* gStdin;
91
92LineReader* open(BigStr* path) {
93 // TODO: Don't use C I/O; use POSIX I/O!
94 FILE* f = fopen(path->data_, "r");
95 if (f == nullptr) {
96 throw Alloc<IOError>(errno);
97 }
98
99 return reinterpret_cast<LineReader*>(Alloc<CFile>(f));
100}
101
102BigStr* CFile::readline() {
103 char* line = nullptr;
104 size_t allocated_size = 0; // unused
105
106 // Reset errno because we turn the EOF error into empty string (like Python).
107 errno = 0;
108 ssize_t len = getline(&line, &allocated_size, f_);
109 if (len < 0) {
110 // man page says the buffer should be freed even if getline fails
111 free(line);
112 if (errno != 0) { // Unexpected error
113 log("getline() error: %s", strerror(errno));
114 throw Alloc<IOError>(errno);
115 }
116 return kEmptyString; // EOF indicated by by empty string, like Python
117 }
118
119 // Note: getline() NUL-terminates the buffer
120 BigStr* result = ::StrFromC(line, len);
121 free(line);
122 return result;
123}
124
125bool CFile::isatty() {
126 return ::isatty(fileno(f_));
127}
128
129// Problem: most BigStr methods like index() and slice() COPY so they have a
130// NUL terminator.
131// log("%s") falls back on sprintf, so it expects a NUL terminator.
132// It would be easier for us to just share.
133BigStr* BufLineReader::readline() {
134 BigStr* line = nullptr;
135
136 int str_len = len(s_);
137 if (pos_ == str_len) {
138 return kEmptyString;
139 }
140
141 int orig_pos = pos_;
142 const char* p = strchr(s_->data_ + pos_, '\n');
143 // log("pos_ = %s", pos_);
144 int line_len;
145 if (p) {
146 int new_pos = p - s_->data_;
147 line_len = new_pos - pos_ + 1; // past newline char
148 pos_ = new_pos + 1;
149 } else { // leftover line
150 if (pos_ == 0) { // The string has no newlines at all -- just return it
151 pos_ = str_len; // advance to the end
152 return s_;
153 } else {
154 line_len = str_len - pos_;
155 pos_ = str_len; // advance to the end
156 }
157 }
158
159 line = NewStr(line_len);
160 memcpy(line->data_, s_->data_ + orig_pos, line_len);
161 DCHECK(line->data_[line_len] == '\0');
162 return line;
163}
164
165Writer* gStdout;
166Writer* gStderr;
167
168//
169// CFileWriter
170//
171
172void CFile::write(BigStr* s) {
173 // Writes can be short!
174 int n = len(s);
175 int num_written = ::fwrite(s->data_, sizeof(char), n, f_);
176 // Similar to CPython fileobject.c
177 if (num_written != n) {
178 throw Alloc<IOError>(errno);
179 }
180}
181
182void CFile::flush() {
183 if (::fflush(f_) != 0) {
184 throw Alloc<IOError>(errno);
185 }
186}
187
188void CFile::close() {
189 if (::fclose(f_) != 0) {
190 throw Alloc<IOError>(errno);
191 }
192}
193
194//
195// BufWriter
196//
197
198void BufWriter::EnsureMoreSpace(int n) {
199 if (str_ == nullptr) {
200 // TODO: we could make the default capacity big enough for a line, e.g. 128
201 // capacity: 128 -> 256 -> 512
202 str_ = NewMutableStr(n);
203 return;
204 }
205
206 int current_cap = len(str_);
207 DCHECK(current_cap >= len_);
208
209 int new_cap = len_ + n;
210
211 if (current_cap < new_cap) {
212 auto* s = NewMutableStr(std::max(current_cap * 2, new_cap));
213 memcpy(s->data_, str_->data_, len_);
214 s->data_[len_] = '\0';
215 str_ = s;
216 }
217}
218
219uint8_t* BufWriter::LengthPointer() {
220 // start + len
221 return reinterpret_cast<uint8_t*>(str_->data_) + len_;
222}
223
224uint8_t* BufWriter::CapacityPointer() {
225 // start + capacity
226 return reinterpret_cast<uint8_t*>(str_->data_) + str_->len_;
227}
228
229void BufWriter::SetLengthFrom(uint8_t* length_ptr) {
230 uint8_t* begin = reinterpret_cast<uint8_t*>(str_->data_);
231 DCHECK(length_ptr >= begin); // we should have written some data
232
233 // Set the length, e.g. so we know where to resume writing from
234 len_ = length_ptr - begin;
235 // printf("SET LEN to %d\n", len_);
236}
237
238void BufWriter::Truncate(int length) {
239 len_ = length;
240}
241
242void BufWriter::WriteRaw(char* s, int n) {
243 DCHECK(is_valid_); // Can't write() after getvalue()
244
245 // write('') is a no-op, so don't create Buf if we don't need to
246 if (n == 0) {
247 return;
248 }
249
250 EnsureMoreSpace(n);
251
252 // Append the contents to the buffer
253 memcpy(str_->data_ + len_, s, n);
254 len_ += n;
255 str_->data_[len_] = '\0';
256}
257
258void BufWriter::WriteConst(const char* c_string) {
259 // meant for short strings like '"'
260 WriteRaw(const_cast<char*>(c_string), strlen(c_string));
261}
262
263void BufWriter::write(BigStr* s) {
264 WriteRaw(s->data_, len(s));
265}
266
267void BufWriter::write_spaces(int n) {
268 if (n == 0) {
269 return;
270 }
271
272 EnsureMoreSpace(n);
273
274 char* dest = str_->data_ + len_;
275 for (int i = 0; i < n; ++i) {
276 dest[i] = ' ';
277 }
278 len_ += n;
279 str_->data_[len_] = '\0';
280}
281
282BigStr* BufWriter::getvalue() {
283 DCHECK(is_valid_); // Check for two INVALID getvalue() in a row
284 is_valid_ = false;
285
286 if (str_ == nullptr) { // if no write() methods are called, the result is ""
287 return kEmptyString;
288 } else {
289 BigStr* s = str_;
290 s->MaybeShrink(len_);
291 str_ = nullptr;
292 len_ = -1;
293 return s;
294 }
295}
296
297} // namespace mylib