1 | #include <fstream>
|
---|
2 | #include <iostream>
|
---|
3 | #include <algorithm>
|
---|
4 | #include <map>
|
---|
5 | #include <vector>
|
---|
6 |
|
---|
7 | #include "Configuration.h"
|
---|
8 |
|
---|
9 | #include "externals/fits.h"
|
---|
10 | #include "externals/huffman.h"
|
---|
11 |
|
---|
12 | #include "Time.h"
|
---|
13 |
|
---|
14 | using namespace std;
|
---|
15 |
|
---|
16 | void SetupConfiguration(Configuration &conf)
|
---|
17 | {
|
---|
18 | po::options_description control("zfits");
|
---|
19 | control.add_options()
|
---|
20 | ("in", var<string>()->required(), "")
|
---|
21 | ("out", var<string>(), "")
|
---|
22 | ("decompress,d", po_switch(), "")
|
---|
23 | ("force,f", var<string>(), "Force overwrite of output file")
|
---|
24 | ;
|
---|
25 |
|
---|
26 | po::positional_options_description p;
|
---|
27 | p.add("in", 1); // The 1st positional options
|
---|
28 | p.add("out", 2); // The 2nd positional options
|
---|
29 |
|
---|
30 | conf.AddOptions(control);
|
---|
31 | conf.SetArgumentPositions(p);
|
---|
32 | }
|
---|
33 |
|
---|
34 | void PrintUsage()
|
---|
35 | {
|
---|
36 | cout <<
|
---|
37 | "zfits - A fits compressor\n"
|
---|
38 | "\n"
|
---|
39 | "\n"
|
---|
40 | "Usage: zfits [-d] input.fits[.gz] [output.zf]\n";
|
---|
41 | cout << endl;
|
---|
42 | }
|
---|
43 |
|
---|
44 |
|
---|
45 | string ReplaceEnd(const string &str, const string &expr, const string &repl)
|
---|
46 | {
|
---|
47 | string out(str);
|
---|
48 |
|
---|
49 | const size_t p = out.rfind(expr);
|
---|
50 | if (p==out.size()-expr.length())
|
---|
51 | out.replace(p, expr.length(), repl);
|
---|
52 |
|
---|
53 | return out;
|
---|
54 | }
|
---|
55 |
|
---|
56 |
|
---|
57 | string ReplaceExt(const string &name, bool decomp)
|
---|
58 | {
|
---|
59 | if (decomp)
|
---|
60 | return ReplaceEnd(name, ".zfits", ".fits");
|
---|
61 |
|
---|
62 | string out = ReplaceEnd(name, ".fits", ".zfits");
|
---|
63 | return ReplaceEnd(out, ".fits.gz", ".zfits");
|
---|
64 | }
|
---|
65 |
|
---|
66 | int Compress(const string &ifile, const string &ofile)
|
---|
67 | {
|
---|
68 | // when to print some info on the screen (every f percent)
|
---|
69 | float frac = 0.01;
|
---|
70 |
|
---|
71 | // open a fits file
|
---|
72 | fits f(ifile);
|
---|
73 |
|
---|
74 | // open output file
|
---|
75 | ofstream fout(ofile);
|
---|
76 |
|
---|
77 | // counters for total size and compressed size
|
---|
78 | uint64_t tot = 0;
|
---|
79 | uint64_t com = 0;
|
---|
80 |
|
---|
81 | // very simple timer
|
---|
82 | double sec = 0;
|
---|
83 |
|
---|
84 | // Produce a lookup table with all informations about the
|
---|
85 | // columns in the same order as they are in the file
|
---|
86 | const fits::Table::Columns &cols= f.GetColumns();
|
---|
87 |
|
---|
88 | struct col_t : fits::Table::Column
|
---|
89 | {
|
---|
90 | string name;
|
---|
91 | void *ptr;
|
---|
92 | };
|
---|
93 |
|
---|
94 |
|
---|
95 | map<size_t, col_t> columns;
|
---|
96 |
|
---|
97 | size_t row_tot = 0;
|
---|
98 | for (auto it=cols.begin(); it!=cols.end(); it++)
|
---|
99 | {
|
---|
100 | col_t c;
|
---|
101 |
|
---|
102 | c.offset = it->second.offset;
|
---|
103 | c.size = it->second.size;
|
---|
104 | c.num = it->second.num;
|
---|
105 | c.name = it->first;
|
---|
106 | c.ptr = f.SetPtrAddress(it->first);
|
---|
107 |
|
---|
108 | columns[c.offset] = c;
|
---|
109 |
|
---|
110 | row_tot += c.size*c.num;
|
---|
111 | }
|
---|
112 |
|
---|
113 | // copy the header from the input to the output file
|
---|
114 | // and prefix the output file as a compressed fits file
|
---|
115 | string header;
|
---|
116 | header.resize(f.tellg());
|
---|
117 |
|
---|
118 | f.seekg(0);
|
---|
119 | f.read((char*)header.c_str(), header.size());
|
---|
120 |
|
---|
121 | char m[2];
|
---|
122 | m[0] = 'z'+128;
|
---|
123 | m[1] = 'f'+128;
|
---|
124 |
|
---|
125 | const size_t hlen = 0;
|
---|
126 |
|
---|
127 | size_t hs = header.size();
|
---|
128 |
|
---|
129 | fout.write(m, 2); // magic number
|
---|
130 | fout.write((char*)&hlen, sizeof(size_t)); // length of possible header data (e.g. file version, compression algorithm)
|
---|
131 | fout.write((char*)&hs, sizeof(size_t)); // size of FITS header
|
---|
132 | fout.write(header.c_str(), header.size()); // uncompressed FITS header
|
---|
133 |
|
---|
134 | tot += header.size();
|
---|
135 | com += header.size()+2+2*sizeof(size_t);
|
---|
136 |
|
---|
137 | cout << fixed;
|
---|
138 |
|
---|
139 | Time start;
|
---|
140 |
|
---|
141 | // loop over all rows
|
---|
142 | vector<char> cache(row_tot);
|
---|
143 | while (f.GetNextRow())
|
---|
144 | {
|
---|
145 | // pointer to the start of the cache for the data of one row
|
---|
146 | char *out = cache.data();
|
---|
147 |
|
---|
148 | // mask stroing which column have been compressed and which not
|
---|
149 | vector<uint8_t> mask(cols.size()/8 + 1);
|
---|
150 |
|
---|
151 | // loop over all columns
|
---|
152 | uint32_t icol = 0;
|
---|
153 | for (auto it=columns.begin(); it!=columns.end(); it++, icol++)
|
---|
154 | {
|
---|
155 | // size of cell in bytes
|
---|
156 | const size_t len_col = it->second.size * it->second.num;
|
---|
157 |
|
---|
158 | // get pointer to data
|
---|
159 | int16_t *ptr = (int16_t*)it->second.ptr;
|
---|
160 |
|
---|
161 | // If the column is the data, preprocess the data
|
---|
162 | /*
|
---|
163 | if (it->second.name=="Data")
|
---|
164 | {
|
---|
165 | int16_t *end = ptr+1440*300-4-(1440*300)%2;
|
---|
166 | int16_t *beg = ptr;
|
---|
167 |
|
---|
168 | while (end>=beg)
|
---|
169 | {
|
---|
170 | const int16_t avg = (end[0] + end[1])/2;
|
---|
171 | end[2] -= avg;
|
---|
172 | end[3] -= avg;
|
---|
173 | end -=2;
|
---|
174 | }
|
---|
175 | }*/
|
---|
176 |
|
---|
177 | // do not try to compress less than 32bytes
|
---|
178 | if (len_col>32 && it->second.size==2)
|
---|
179 | {
|
---|
180 | Time now;
|
---|
181 |
|
---|
182 | // perform 16bit hoffman (option for 8bit missing, skip 64bit)
|
---|
183 | // (what to do with floats?)
|
---|
184 | string buf;
|
---|
185 | /*int len =*/ Huffman::Encode(buf, (uint16_t*)ptr, len_col/2);
|
---|
186 |
|
---|
187 | sec += Time().UnixTime()-now.UnixTime();
|
---|
188 |
|
---|
189 | // check if data was really compressed
|
---|
190 | if (buf.size()<len_col)
|
---|
191 | {
|
---|
192 | // copy compressed data into output cache
|
---|
193 | memcpy(out, buf.c_str(), buf.size());
|
---|
194 | out += buf.size();
|
---|
195 |
|
---|
196 | // update mask
|
---|
197 | const uint64_t bit = (icol%8);
|
---|
198 | mask[icol/8] |= (1<<bit);
|
---|
199 |
|
---|
200 | continue;
|
---|
201 | }
|
---|
202 | }
|
---|
203 |
|
---|
204 | // just copy the data if it has not been compressed
|
---|
205 | memcpy(out, (char*)ptr, len_col);
|
---|
206 | out += len_col;
|
---|
207 | }
|
---|
208 |
|
---|
209 | // calcualte size of output buffer
|
---|
210 | const size_t sz = out-cache.data();
|
---|
211 |
|
---|
212 | // update counters
|
---|
213 | tot += row_tot;
|
---|
214 | com += sz + mask.size();
|
---|
215 |
|
---|
216 | // write the compression mask and the (partly) copmpressed data stream
|
---|
217 | fout.write((char*)mask.data(), mask.size());
|
---|
218 | fout.write(cache.data(), sz);
|
---|
219 |
|
---|
220 | //if (sz2<0 || memcmp(data, dest3.data(), 432000*2)!=0)
|
---|
221 | // cout << "grrrr" << endl;
|
---|
222 |
|
---|
223 | const float proc = float(f.GetRow())/f.GetNumRows();
|
---|
224 | if (proc>frac)
|
---|
225 | {
|
---|
226 | const double elep = Time().UnixTime()-start.UnixTime();
|
---|
227 | cout << "\r" << setprecision(0) << setw(3) << 100*proc << "% [" << setprecision(1) << setw(5) << 100.*com/tot << "%] cpu:" << sec << "s in:" << tot/1000000/elep << "MB/s" << flush;
|
---|
228 | frac += 0.01;
|
---|
229 | }
|
---|
230 | }
|
---|
231 |
|
---|
232 | const double elep = Time().UnixTime()-start.UnixTime();
|
---|
233 | cout << setprecision(0) << "\r100% [" << setprecision(1) << setw(5) << 100.*com/tot << "%] cpu:" << sec << "s in:" << tot/1000000/elep << "MB/s" << endl;
|
---|
234 |
|
---|
235 | return 0;
|
---|
236 | }
|
---|
237 |
|
---|
238 | template<size_t N>
|
---|
239 | void revcpy(char *dest, const char *src, int num)
|
---|
240 | {
|
---|
241 | const char *pend = src + num*N;
|
---|
242 | for (const char *ptr = src; ptr<pend; ptr+=N, dest+=N)
|
---|
243 | reverse_copy(ptr, ptr+N, dest);
|
---|
244 | }
|
---|
245 |
|
---|
246 | int Decompress(const string &ifile, const string &ofile)
|
---|
247 | {
|
---|
248 | // open a fits file
|
---|
249 | ifstream fin(ifile);
|
---|
250 |
|
---|
251 | // open output file
|
---|
252 | ofstream fout(ofile);
|
---|
253 |
|
---|
254 | // get and check magic number
|
---|
255 | unsigned char m[2];
|
---|
256 | fin.read((char*)m, 2);
|
---|
257 | if (m[0]!='z'+128 || m[1]!='f'+128)
|
---|
258 | throw runtime_error("File not a compressed fits file.");
|
---|
259 |
|
---|
260 | // get length of additional header information
|
---|
261 | size_t hlen = 0;
|
---|
262 | fin.read((char*)&hlen, sizeof(size_t));
|
---|
263 | if (hlen>0)
|
---|
264 | throw runtime_error("Only Version-zero files supported.");
|
---|
265 |
|
---|
266 | // get size of FITS header
|
---|
267 | size_t hs = 0;
|
---|
268 | fin.read((char*)&hs, sizeof(size_t));
|
---|
269 | if (!fin)
|
---|
270 | throw runtime_error("Could not access header size.");
|
---|
271 |
|
---|
272 | // copy the header from the input to the output file
|
---|
273 | // and prefix the output file as a compressed fits file
|
---|
274 | string header;
|
---|
275 | header.resize(hs);
|
---|
276 |
|
---|
277 | fin.read((char*)header.c_str(), header.size());
|
---|
278 | fout.write((char*)header.c_str(), header.size());
|
---|
279 | if (!fin)
|
---|
280 | throw runtime_error("Could not read full header");
|
---|
281 |
|
---|
282 | string templ("tmpXXXXXX");
|
---|
283 | int fd = mkstemp((char*)templ.c_str());
|
---|
284 | const ssize_t rc = write(fd, header.c_str(), header.size());
|
---|
285 | close(fd);
|
---|
286 |
|
---|
287 | if (rc<0)
|
---|
288 | throw runtime_error("Could not write to temporary file: "+string(strerror(errno)));
|
---|
289 |
|
---|
290 | // open the output file to get the header parsed
|
---|
291 | fits info(templ);
|
---|
292 |
|
---|
293 | remove(templ.c_str());
|
---|
294 |
|
---|
295 | // get the maximum size of one row and a list
|
---|
296 | // of all columns ordered by their offset
|
---|
297 | size_t row_tot = 0;
|
---|
298 | const fits::Table::Columns &cols = info.GetColumns();
|
---|
299 | map<size_t, fits::Table::Column> columns;
|
---|
300 | for (auto it=cols.begin(); it!=cols.end(); it++)
|
---|
301 | {
|
---|
302 | columns[it->second.offset] = it->second;
|
---|
303 | row_tot += it->second.num*it->second.size;
|
---|
304 | }
|
---|
305 |
|
---|
306 | // very simple timer
|
---|
307 | double sec = 0;
|
---|
308 | double frac = 0;
|
---|
309 |
|
---|
310 | size_t com = 2+hs+2*sizeof(size_t);
|
---|
311 | size_t tot = hs;
|
---|
312 |
|
---|
313 | const size_t masklen = cols.size()/8+1;
|
---|
314 |
|
---|
315 | // loop over all rows
|
---|
316 | vector<char> buf(row_tot+masklen);
|
---|
317 | vector<char> swap(row_tot);
|
---|
318 | uint32_t offset = 0;
|
---|
319 |
|
---|
320 | const uint64_t nrows = info.GetUInt("NAXIS2");
|
---|
321 |
|
---|
322 | Time start;
|
---|
323 |
|
---|
324 | cout << fixed;
|
---|
325 | for (uint32_t irow=0; irow<nrows; irow++)
|
---|
326 | {
|
---|
327 | fin.read(buf.data()+offset, buf.size()-offset);
|
---|
328 |
|
---|
329 | const uint8_t *mask = reinterpret_cast<uint8_t*>(buf.data());
|
---|
330 | offset = masklen;
|
---|
331 |
|
---|
332 | char *ptr = swap.data();
|
---|
333 |
|
---|
334 | uint32_t icol = 0;
|
---|
335 | for (auto it=columns.begin(); it!= columns.end(); it++, icol++)
|
---|
336 | {
|
---|
337 | const size_t &num = it->second.num;
|
---|
338 | const size_t &size = it->second.size;
|
---|
339 |
|
---|
340 | if (mask[icol/8]&(1<<(icol%8)))
|
---|
341 | {
|
---|
342 | Time now;
|
---|
343 |
|
---|
344 | vector<uint16_t> out(num*size/2);
|
---|
345 | int len = Huffman::Decode((uint8_t*)buf.data()+offset, buf.size()-offset, out);
|
---|
346 | if (len<0)
|
---|
347 | throw runtime_error("Decoding failed.");
|
---|
348 |
|
---|
349 | sec += Time().UnixTime()-now.UnixTime();
|
---|
350 |
|
---|
351 | offset += len;
|
---|
352 |
|
---|
353 | revcpy<2>(ptr, (char*)out.data(), num);
|
---|
354 | }
|
---|
355 | else
|
---|
356 | {
|
---|
357 | switch (size)
|
---|
358 | {
|
---|
359 | case 1: memcpy (ptr, buf.data()+offset, num*size); break;
|
---|
360 | case 2: revcpy<2>(ptr, buf.data()+offset, num); break;
|
---|
361 | case 4: revcpy<4>(ptr, buf.data()+offset, num); break;
|
---|
362 | case 8: revcpy<8>(ptr, buf.data()+offset, num); break;
|
---|
363 | }
|
---|
364 |
|
---|
365 | offset += num*size;
|
---|
366 | }
|
---|
367 |
|
---|
368 | ptr += num*size;
|
---|
369 | }
|
---|
370 |
|
---|
371 | com += offset+masklen;
|
---|
372 | tot += row_tot;
|
---|
373 |
|
---|
374 | fout.write((char*)swap.data(), swap.size());
|
---|
375 |
|
---|
376 | memmove(buf.data(), buf.data()+offset, buf.size()-offset);
|
---|
377 | offset = buf.size()-offset;
|
---|
378 |
|
---|
379 | if (!fout)
|
---|
380 | throw runtime_error("Error writing to output file");
|
---|
381 |
|
---|
382 | const float proc = float(irow)/nrows;
|
---|
383 | if (proc>frac)
|
---|
384 | {
|
---|
385 | const double elep = Time().UnixTime()-start.UnixTime();
|
---|
386 | cout << "\r" << setprecision(0) << setw(3) << 100*proc << "% [" << setprecision(1) << setw(5) << 100.*com/tot << "%] cpu:" << sec << "s out:" << tot/1000000/elep << "MB/s" << flush;
|
---|
387 | frac += 0.01;
|
---|
388 | }
|
---|
389 | }
|
---|
390 |
|
---|
391 | const double elep = Time().UnixTime()-start.UnixTime();
|
---|
392 | cout << setprecision(0) << "\r100% [" << setprecision(1) << setw(5) << 100.*com/tot << "%] cpu:" << sec << "s out:" << tot/1000000/elep << "MB/s" << endl;
|
---|
393 |
|
---|
394 | return 0;
|
---|
395 | }
|
---|
396 |
|
---|
397 | int main(int argc, const char **argv)
|
---|
398 | {
|
---|
399 | Configuration conf(argv[0]);
|
---|
400 | conf.SetPrintUsage(PrintUsage);
|
---|
401 | SetupConfiguration(conf);
|
---|
402 |
|
---|
403 | if (!conf.DoParse(argc, argv))
|
---|
404 | return 127;
|
---|
405 |
|
---|
406 | const bool decomp = conf.Get<bool>("decompress");
|
---|
407 |
|
---|
408 | const string ifile = conf.Get<string>("in");
|
---|
409 | const string ofile = conf.Has("out") ? conf.Get<string>("out") : ReplaceExt(ifile, decomp);
|
---|
410 |
|
---|
411 | return decomp ? Decompress(ifile, ofile) : Compress(ifile, ofile);
|
---|
412 |
|
---|
413 | /*
|
---|
414 | // reading and writing files which just contain the binary data
|
---|
415 | // For simplicity I assume ROI=300
|
---|
416 |
|
---|
417 | ifstream finx( "20130117_082.fits.pu");
|
---|
418 | ofstream foutx("20130117_082.fits.puz");
|
---|
419 |
|
---|
420 | while (1)
|
---|
421 | {
|
---|
422 | string str;
|
---|
423 | str.resize(432000*2);
|
---|
424 |
|
---|
425 | finx.read((char*)str.c_str(), 432000*2);
|
---|
426 | if (!finx)
|
---|
427 | break;
|
---|
428 |
|
---|
429 | // Preprocess the data, e.g. subtract median pixelwise
|
---|
430 | for (int i=0; i<1440; i++)
|
---|
431 | {
|
---|
432 | int16_t *chunk = (int16_t*)str.c_str()+i*300;
|
---|
433 | sort(chunk, chunk+300);
|
---|
434 |
|
---|
435 | int16_t med = chunk[149];
|
---|
436 |
|
---|
437 | for (int j=0; j<300; j++)
|
---|
438 | chunk[j] -= med;
|
---|
439 | }
|
---|
440 |
|
---|
441 | // do huffman encoding on shorts
|
---|
442 | string buf;
|
---|
443 | int len = huffmans_encode(buf, (uint16_t*)str.c_str(), 432000);
|
---|
444 |
|
---|
445 | // if the result is smaller than the original data write
|
---|
446 | // the result, otherwise the original data
|
---|
447 | if (buf.size()<432000*2)
|
---|
448 | foutx.write(buf.c_str(), buf.size());
|
---|
449 | else
|
---|
450 | foutx.write(str.c_str(), 432000);
|
---|
451 | }
|
---|
452 |
|
---|
453 | return 0;
|
---|
454 | */
|
---|
455 | }
|
---|