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

300 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 // Reset EOF flag so the next readline() will get the line
116 clearerr(f_);
117 throw Alloc<IOError>(errno);
118 }
119 return kEmptyString; // EOF indicated by by empty string, like Python
120 }
121
122 // Note: getline() NUL-terminates the buffer
123 BigStr* result = ::StrFromC(line, len);
124 free(line);
125 return result;
126}
127
128bool CFile::isatty() {
129 return ::isatty(fileno(f_));
130}
131
132// Problem: most BigStr methods like index() and slice() COPY so they have a
133// NUL terminator.
134// log("%s") falls back on sprintf, so it expects a NUL terminator.
135// It would be easier for us to just share.
136BigStr* BufLineReader::readline() {
137 BigStr* line = nullptr;
138
139 int str_len = len(s_);
140 if (pos_ == str_len) {
141 return kEmptyString;
142 }
143
144 int orig_pos = pos_;
145 const char* p = strchr(s_->data_ + pos_, '\n');
146 // log("pos_ = %s", pos_);
147 int line_len;
148 if (p) {
149 int new_pos = p - s_->data_;
150 line_len = new_pos - pos_ + 1; // past newline char
151 pos_ = new_pos + 1;
152 } else { // leftover line
153 if (pos_ == 0) { // The string has no newlines at all -- just return it
154 pos_ = str_len; // advance to the end
155 return s_;
156 } else {
157 line_len = str_len - pos_;
158 pos_ = str_len; // advance to the end
159 }
160 }
161
162 line = NewStr(line_len);
163 memcpy(line->data_, s_->data_ + orig_pos, line_len);
164 DCHECK(line->data_[line_len] == '\0');
165 return line;
166}
167
168Writer* gStdout;
169Writer* gStderr;
170
171//
172// CFileWriter
173//
174
175void CFile::write(BigStr* s) {
176 // Writes can be short!
177 int n = len(s);
178 int num_written = ::fwrite(s->data_, sizeof(char), n, f_);
179 // Similar to CPython fileobject.c
180 if (num_written != n) {
181 throw Alloc<IOError>(errno);
182 }
183}
184
185void CFile::flush() {
186 if (::fflush(f_) != 0) {
187 throw Alloc<IOError>(errno);
188 }
189}
190
191void CFile::close() {
192 if (::fclose(f_) != 0) {
193 throw Alloc<IOError>(errno);
194 }
195}
196
197//
198// BufWriter
199//
200
201void BufWriter::EnsureMoreSpace(int n) {
202 if (str_ == nullptr) {
203 // TODO: we could make the default capacity big enough for a line, e.g. 128
204 // capacity: 128 -> 256 -> 512
205 str_ = NewMutableStr(n);
206 return;
207 }
208
209 int current_cap = len(str_);
210 DCHECK(current_cap >= len_);
211
212 int new_cap = len_ + n;
213
214 if (current_cap < new_cap) {
215 auto* s = NewMutableStr(std::max(current_cap * 2, new_cap));
216 memcpy(s->data_, str_->data_, len_);
217 s->data_[len_] = '\0';
218 str_ = s;
219 }
220}
221
222uint8_t* BufWriter::LengthPointer() {
223 // start + len
224 return reinterpret_cast<uint8_t*>(str_->data_) + len_;
225}
226
227uint8_t* BufWriter::CapacityPointer() {
228 // start + capacity
229 return reinterpret_cast<uint8_t*>(str_->data_) + str_->len_;
230}
231
232void BufWriter::SetLengthFrom(uint8_t* length_ptr) {
233 uint8_t* begin = reinterpret_cast<uint8_t*>(str_->data_);
234 DCHECK(length_ptr >= begin); // we should have written some data
235
236 // Set the length, e.g. so we know where to resume writing from
237 len_ = length_ptr - begin;
238 // printf("SET LEN to %d\n", len_);
239}
240
241void BufWriter::Truncate(int length) {
242 len_ = length;
243}
244
245void BufWriter::WriteRaw(char* s, int n) {
246 DCHECK(is_valid_); // Can't write() after getvalue()
247
248 // write('') is a no-op, so don't create Buf if we don't need to
249 if (n == 0) {
250 return;
251 }
252
253 EnsureMoreSpace(n);
254
255 // Append the contents to the buffer
256 memcpy(str_->data_ + len_, s, n);
257 len_ += n;
258 str_->data_[len_] = '\0';
259}
260
261void BufWriter::WriteConst(const char* c_string) {
262 // meant for short strings like '"'
263 WriteRaw(const_cast<char*>(c_string), strlen(c_string));
264}
265
266void BufWriter::write(BigStr* s) {
267 WriteRaw(s->data_, len(s));
268}
269
270void BufWriter::write_spaces(int n) {
271 if (n == 0) {
272 return;
273 }
274
275 EnsureMoreSpace(n);
276
277 char* dest = str_->data_ + len_;
278 for (int i = 0; i < n; ++i) {
279 dest[i] = ' ';
280 }
281 len_ += n;
282 str_->data_[len_] = '\0';
283}
284
285BigStr* BufWriter::getvalue() {
286 DCHECK(is_valid_); // Check for two INVALID getvalue() in a row
287 is_valid_ = false;
288
289 if (str_ == nullptr) { // if no write() methods are called, the result is ""
290 return kEmptyString;
291 } else {
292 BigStr* s = str_;
293 s->MaybeShrink(len_);
294 str_ = nullptr;
295 len_ = -1;
296 return s;
297 }
298}
299
300} // namespace mylib