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

298 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 // log("getline = %d", len);
110 if (len < 0) {
111 // man page says the buffer should be freed even if getline fails
112 free(line);
113 if (errno != 0) { // Unexpected error
114 log("getline() error: %s", strerror(errno));
115 throw Alloc<IOError>(errno);
116 }
117 return kEmptyString; // EOF indicated by by empty string, like Python
118 }
119
120 // Note: getline() NUL-terminates the buffer
121 BigStr* result = ::StrFromC(line, len);
122 free(line);
123 return result;
124}
125
126bool CFile::isatty() {
127 return ::isatty(fileno(f_));
128}
129
130// Problem: most BigStr methods like index() and slice() COPY so they have a
131// NUL terminator.
132// log("%s") falls back on sprintf, so it expects a NUL terminator.
133// It would be easier for us to just share.
134BigStr* BufLineReader::readline() {
135 BigStr* line = nullptr;
136
137 int str_len = len(s_);
138 if (pos_ == str_len) {
139 return kEmptyString;
140 }
141
142 int orig_pos = pos_;
143 const char* p = strchr(s_->data_ + pos_, '\n');
144 // log("pos_ = %s", pos_);
145 int line_len;
146 if (p) {
147 int new_pos = p - s_->data_;
148 line_len = new_pos - pos_ + 1; // past newline char
149 pos_ = new_pos + 1;
150 } else { // leftover line
151 if (pos_ == 0) { // The string has no newlines at all -- just return it
152 pos_ = str_len; // advance to the end
153 return s_;
154 } else {
155 line_len = str_len - pos_;
156 pos_ = str_len; // advance to the end
157 }
158 }
159
160 line = NewStr(line_len);
161 memcpy(line->data_, s_->data_ + orig_pos, line_len);
162 DCHECK(line->data_[line_len] == '\0');
163 return line;
164}
165
166Writer* gStdout;
167Writer* gStderr;
168
169//
170// CFileWriter
171//
172
173void CFile::write(BigStr* s) {
174 // Writes can be short!
175 int n = len(s);
176 int num_written = ::fwrite(s->data_, sizeof(char), n, f_);
177 // Similar to CPython fileobject.c
178 if (num_written != n) {
179 throw Alloc<IOError>(errno);
180 }
181}
182
183void CFile::flush() {
184 if (::fflush(f_) != 0) {
185 throw Alloc<IOError>(errno);
186 }
187}
188
189void CFile::close() {
190 if (::fclose(f_) != 0) {
191 throw Alloc<IOError>(errno);
192 }
193}
194
195//
196// BufWriter
197//
198
199void BufWriter::EnsureMoreSpace(int n) {
200 if (str_ == nullptr) {
201 // TODO: we could make the default capacity big enough for a line, e.g. 128
202 // capacity: 128 -> 256 -> 512
203 str_ = NewMutableStr(n);
204 return;
205 }
206
207 int current_cap = len(str_);
208 DCHECK(current_cap >= len_);
209
210 int new_cap = len_ + n;
211
212 if (current_cap < new_cap) {
213 auto* s = NewMutableStr(std::max(current_cap * 2, new_cap));
214 memcpy(s->data_, str_->data_, len_);
215 s->data_[len_] = '\0';
216 str_ = s;
217 }
218}
219
220uint8_t* BufWriter::LengthPointer() {
221 // start + len
222 return reinterpret_cast<uint8_t*>(str_->data_) + len_;
223}
224
225uint8_t* BufWriter::CapacityPointer() {
226 // start + capacity
227 return reinterpret_cast<uint8_t*>(str_->data_) + str_->len_;
228}
229
230void BufWriter::SetLengthFrom(uint8_t* length_ptr) {
231 uint8_t* begin = reinterpret_cast<uint8_t*>(str_->data_);
232 DCHECK(length_ptr >= begin); // we should have written some data
233
234 // Set the length, e.g. so we know where to resume writing from
235 len_ = length_ptr - begin;
236 // printf("SET LEN to %d\n", len_);
237}
238
239void BufWriter::Truncate(int length) {
240 len_ = length;
241}
242
243void BufWriter::WriteRaw(char* s, int n) {
244 DCHECK(is_valid_); // Can't write() after getvalue()
245
246 // write('') is a no-op, so don't create Buf if we don't need to
247 if (n == 0) {
248 return;
249 }
250
251 EnsureMoreSpace(n);
252
253 // Append the contents to the buffer
254 memcpy(str_->data_ + len_, s, n);
255 len_ += n;
256 str_->data_[len_] = '\0';
257}
258
259void BufWriter::WriteConst(const char* c_string) {
260 // meant for short strings like '"'
261 WriteRaw(const_cast<char*>(c_string), strlen(c_string));
262}
263
264void BufWriter::write(BigStr* s) {
265 WriteRaw(s->data_, len(s));
266}
267
268void BufWriter::write_spaces(int n) {
269 if (n == 0) {
270 return;
271 }
272
273 EnsureMoreSpace(n);
274
275 char* dest = str_->data_ + len_;
276 for (int i = 0; i < n; ++i) {
277 dest[i] = ' ';
278 }
279 len_ += n;
280 str_->data_[len_] = '\0';
281}
282
283BigStr* BufWriter::getvalue() {
284 DCHECK(is_valid_); // Check for two INVALID getvalue() in a row
285 is_valid_ = false;
286
287 if (str_ == nullptr) { // if no write() methods are called, the result is ""
288 return kEmptyString;
289 } else {
290 BigStr* s = str_;
291 s->MaybeShrink(len_);
292 str_ = nullptr;
293 len_ = -1;
294 return s;
295 }
296}
297
298} // namespace mylib