File Sink: Difference between revisions
Jesternofool (talk | contribs) (Corrected format for integer. It IS two's complement.) |
|||
(2 intermediate revisions by the same user not shown) | |||
Line 25: | Line 25: | ||
=== What is the file format of a file_sink? How can I read files produced by a file sink? === | === What is the file format of a file_sink? How can I read files produced by a file sink? === | ||
All files written by [[File_Sink|File Sink]] are in pure binary format with no metadata. | All files written by [[File_Sink|File Sink]] are in pure binary format with no metadata. The specific format is dependent upon the specific sample types used to save the data. The possible data types and their storage are: | ||
* complex: two 32-bit floating point numbers, one each for the real and imaginary components. The floating point numbers conform to the [https://en.wikipedia.org/wiki/IEEE_754 IEEE 754 specification]. The output file will require 8 bytes for each sample. The real and imaginary are saved in alternating fashion (real-imag-real-imag...). Reading back a complex number means reading in 32 bits, saving that to the real part of a complex data structure, and then reading in the next 32 bits as the imaginary part of the data structure. And just keep reading the data. | |||
* float: one 32-bit floating point number. The floating point number conforms to the [https://en.wikipedia.org/wiki/IEEE_754 IEEE 754 specification]. The output file will require 4 bytes for each sample. | |||
* int (integer): one 32-bit signed integer stored in [https://en.wikipedia.org/wiki/Two%27s_complement two's complement format]. It can represent integers from -2<sup>31</sup> - +2<sup>31</sup>-1. The output file will require 4 bytes for each sample. | |||
* short: one 16-bit signed integer stored in [https://en.wikipedia.org/wiki/Two%27s_complement two's complement format]. It can represent integers from -32768 to + 32767. The output file will require 2 bytes for each sample. | |||
* byte: one 8-bit signed integer stored in [https://en.wikipedia.org/wiki/Two%27s_complement two's complement format]. It can represent integers from -128 to +127. The output file will require 1 byte for each sample. | |||
The exception to the format is when using the metadata file format. These files are produced by the [[File_Meta_Sink|File Meta Sink]] block and read by the [[File_Meta_Source|File Meta Source]] block. See the manual page on the [[Metadata_Information|Metadata Information]] for more information about how to deal with these files. | The exception to the format is when using the metadata file format. These files are produced by the [[File_Meta_Sink|File Meta Sink]] block and read by the [[File_Meta_Source|File Meta Source]] block. See the manual page on the [[Metadata_Information|Metadata Information]] for more information about how to deal with these files. |
Latest revision as of 01:24, 3 May 2024
Used to write a stream to a binary file.
This file can be read into any programming environment that can read binary files (MATLAB, C, Python, ...). It can also be played back in GRC using a File Source. For example, if complex type is chosen, then the binary file will be full of float32s in IQIQIQ order. There is no meta data or anything else included with the binary data. For more information on handling this data, see the Handling File Sink data section below.
Parameters
(R): Run-time adjustable
- File (R)
- Path of the file to open and write output to. If the specified file name does not exist at that location, it creates a file of that name over there. Otherwise, if the file already exists, it may overwrite or append the file based on the append option.
- Unbuffered
- Specifies whether the output is buffered in memory. If the output is unbuffered, the data will be flushed to the file each time the work function is called. This can cause the flowgraph to run slow due to the time required to access the disk each time.
- Append File
- Gives an option to either append to the file or to overwrite the file.
Example Flowgraph
This flowgraph shows a File Sink which outputs text to the terminal (/dev/stdout).
Handling File Sink data
What is the file format of a file_sink? How can I read files produced by a file sink?
All files written by File Sink are in pure binary format with no metadata. The specific format is dependent upon the specific sample types used to save the data. The possible data types and their storage are:
- complex: two 32-bit floating point numbers, one each for the real and imaginary components. The floating point numbers conform to the IEEE 754 specification. The output file will require 8 bytes for each sample. The real and imaginary are saved in alternating fashion (real-imag-real-imag...). Reading back a complex number means reading in 32 bits, saving that to the real part of a complex data structure, and then reading in the next 32 bits as the imaginary part of the data structure. And just keep reading the data.
- float: one 32-bit floating point number. The floating point number conforms to the IEEE 754 specification. The output file will require 4 bytes for each sample.
- int (integer): one 32-bit signed integer stored in two's complement format. It can represent integers from -231 - +231-1. The output file will require 4 bytes for each sample.
- short: one 16-bit signed integer stored in two's complement format. It can represent integers from -32768 to + 32767. The output file will require 2 bytes for each sample.
- byte: one 8-bit signed integer stored in two's complement format. It can represent integers from -128 to +127. The output file will require 1 byte for each sample.
The exception to the format is when using the metadata file format. These files are produced by the File Meta Sink block and read by the File Meta Source block. See the manual page on the Metadata Information for more information about how to deal with these files.
Reading from Python
A one-line Python command to read the entire file into a numpy array is:
import numpy
f = numpy.fromfile(open("filename"), dtype=)
Where dtype is one of numpy.int16
, numpy.int32
, numpy.float32
, numpy.complex64
or whatever type you were using.
Reading from Octave / Matlab
If you're using octave (or are stuck with using Matlab, our condolences), the following is a method to read all floats written with a File Sink to a variable:
f = fopen('filename', 'rb');
values = fread(f, Inf, 'float');
Replace 'float'
with 'short'
,'int'
or 'char'
as appropriate.
Use complex_v = values(1:2:end) + values(2:2:end)*i;
to convert interleaved real, imaginary values to an array of complex values.
Reading from plain C
This is the easiest, since we assume you know how `fread` works if you're writing C, and how to cast a pointer:
// this is C99.
#include <math.h>
#include <stdio.h>
#include <stdlib.h>
// set the type of data you want to read here
typedef float sample_type;
int main(int argc, char **argv) {
if (argc != 2) {
fputs("Expected one argument!\n", stderr);
exit(-2);
}
FILE *fp = fopen(argv[1], "rb");
if (!fp) {
perror("Error opening file");
exit(-1);
}
// allocate a buffer for 1024 samples
const unsigned int buf_length = 1024;
sample_type *buffer = malloc(buf_length * sizeof(sample_type));
// loop until we don't
while (1) {
// try to read as many samples as fit the buffer
size_t read_count = fread(buffer, sizeof(sample_type), buf_length, fp);
// check for end-of-file / error
if (!read_count) {
break;
}
for (size_t index = 0; index < read_count; ++index) {
// Do something to each sample
double absolute = fabs(buffer[index]);
printf("The absolute of the sample is %f\n", absolute);
}
}
free(buffer);
fclose(fp);
}
Furthermore, here's a way of simply mapping your file to memory and use that to access the values, directly. This is sensible if you do not intend to load the full file into RAM, but want to jump around in it.
// this is C99.
#include <sys/mman.h>
#include <sys/stat.h>
#include <stdio.h>
#include <complex.h>
…
int fd = open("filename", "rb"); // please check fd != -1
struct stat stat_struct;
int retval = fstat(fd, &stat_struct); // please check for ==0
size_t filesize = stat_struct.st_size;
char* address = mmap(NULL, filesize, PROT_READ, MAP_PRIVATE, fd, 0); // please check for address != MAP_FAILEP
// interpret as pointer to a complex float value (could alternatively use float*, int*, ...
complex float* complex_values = (complex float*) address;
// do stuff on complex_values[i]...
…
// finally, clean up:
munmap(address, filesize);
close(fd);
Reading from C++
There's many ways to do this in C++. In this example, we're reading a file, either at once into a large buffer, or in chunks of limited size, until we're done. We're using std::ifstream
as our means of reading the file. This example is rather complete, and illustrates a lot of error handling, file checking, as well as chunk-wise reading etc. It also calculates instantaneous powers, and it outputs these. Your own implementation might be shorter!
// This is C++17
#include <algorithm>
#include <cmath>
#include <complex>
#include <cstddef>
#include <filesystem>
#include <fstream>
#include <string_view>
#include <vector>
#include <fmt/format.h>
#include <fmt/ranges.h>
using sample_t = std::complex<float>;
using power_t = float;
constexpr std::size_t read_block_size = 1 << 16;
int main(int argc, char *argv[]) {
// expect exactly one argument, a file name
if (argc != 2) {
fmt::print(stderr, "Usage: {} FILE_NAME", argv[0]);
return -1;
}
// just for convenience; we could as well just use `argv[1]` throughout the
// code
std::string_view filename(argv[1]);
// check whether file exists
if (!std::filesystem::exists(filename.data())) {
fmt::print(stderr, "file '{:s}' not found\n", filename);
return -2;
}
// calculate how many samples to read
auto file_size = std::filesystem::file_size(std::filesystem::path(filename));
auto samples_to_read = file_size / sizeof(sample_t);
// construct and reserve container for resulting powers
std::vector<power_t> powers;
powers.reserve(samples_to_read);
std::ifstream input_file(filename.data(), std::ios_base::binary);
if (!input_file) {
fmt::print(stderr, "error opening '{:s}'\n", filename);
return -3;
}
// construct and reserve container for read samples
// if read_block_size == 0, then read the whole file at once
std::vector<sample_t> samples;
if (read_block_size)
samples.resize(read_block_size);
else
samples.resize(samples_to_read);
fmt::print(stderr, "Reading {:d} samples…\n", samples_to_read);
while (samples_to_read) {
auto read_now = std::min(samples_to_read, samples.size());
input_file.read(reinterpret_cast<char *>(samples.data()),
read_now * sizeof(sample_t));
for (size_t idx = 0; idx < read_now; ++idx) {
auto magnitude = std::abs(samples[idx]);
powers.push_back(magnitude * magnitude);
}
samples_to_read -= read_now;
}
// we're not actually doing anything with the data. Let's print it!
fmt::print("Power\n{}\n", fmt::join(powers, "\n"));
}
Source Files
- C++ files
- file_sink_impl.cc
- Header files
- file_sink_impl.h
- Public header files
- file_sink.h
- Block definition
- blocks_file_sink.block.yml