fapyc 0.6.0

A Python wrapper for the FAPEC data compressor.

FaPyc

A Python wrapper for the FAPEC data compressor. (C) DAPCOM Data Services S.L. - https://www.dapcom.es

The full FAPEC compression and decompression library is included in this package, but a valid license file must be available to properly use it. Without a license, you can still use the decompressor (yet with some limitations, such as the maximum number of threads, the recovery of corrupted files, or the decompression of just one part of a multi-part archive). You can get free evaluation licenses at https://www.dapcom.es/get-fapec/ to test the compressor. For full licenses, please contact us at [e-mail] Once a valid license is obtained (either full or evaluation), you must define a FAPEC_HOME environment variable pointing to the path where you have stored your fapeclic.dat license file.

Usage

There are 3 main execution modes:

• **File: When invoking Fapyc or Unfapyc on a filename, it will (de)compress it directly into another file.
• Buffer: You can load the whole file to (de)compress on e.g. a byte array, and then invoke Fapyc/Unfapyc which will leave the result in the output buffer. Obviously, you should be careful with large files, as it may use a lot of RAM.
• Chunk: FAPEC internally works in 'chunks' of data, typically 1-8 MB each (maximum 384MB each), which allows to progressively (de)compress a huge file while keeping memory usage under control. File and buffer (de)compression automatically uses this feature. For now, directly invoking this method is only available in the native C API, not in fapyc yet.

The file and buffer operations can also be combined:

• Buffer-to-file compression: You can pass a buffer to Fapyc and tell it to progressively compress and store it into a file.
• File-to-buffer decompression: You can directly decompress a file (without having to load it beforehand) and leave its decompressed output in a buffer, which you can use afterwards.

The basic syntax for these different modes is as follows:

• File-to-file compression:

    from fapyc import Fapyc
    f = Fapyc(filename = your_file)
    f.compress_auto(output = your_file + ".fapec")  # We can also invoke a specific compression algorithm

• Buffer-to-file compression:

    from fapyc import Fapyc
    f = Fapyc(buffer = your_data_buffer)
    f.compress_auto(output = "your_output_file.fapec")

• Buffer-to-buffer compression:

    from fapyc import Fapyc
    f = Fapyc(buffer = your_data_buffer)
    f.compress_auto()
    your_data_handling_routine(f.outputBuffer)

• File-to-file decompression:

    from fapyc import Unfapyc
    uf = Unfapyc(filename = your_fapec_file)
    uf.decompress(output = your_fapec_file + ".restored")  # or whatever filename/extension

• File-to-buffer decompression:

    from fapyc import Unfapyc
    uf = Unfapyc(filename = your_fapec_file)
    uf.decompress()
    your_data_handling_routine(uf.outputBuffer)

• Buffer-to-buffer decompression:

    from fapyc import Unfapyc
    uf = Unfapyc(buffer = your_data_buffer)
    uf.decompress()
    your_data_handling_routine(uf.outputBuffer)

In the current fapyc version, the following compression algorithms and parameters are available:

• Automatic selection of the compression algorithm from the data contents:

  compress_auto()

• LZW dictionary coding:

  compress_lzw()

• Basic integer compression, allowing to indicate the bits per sample, signed integers (True/False), big endian (True/False), interleaving in samples, and lossy level:

  compress_basic(bits, sign, bigendian, il, lossy)

• Tabulated text compression, allowing to indicate the separator character (and even a second separator):

  compress_tabtxt(sep1, sep2)

• Double-precision floating point values, with interleaving and lossy level:

  compress_doubles(bigEndian, il, lossy)

• FastQ genomic files compression:

  compress_fastq()

• Kongsberg's .all files:

  compress_kall()

• Kongsberg's .wcd files:

  compress_kwcd(lossy)

• Kongsberg's .kmall and .kmwcd files:

  compress_kmall(sndlossy, silossy, amplossy, phaselossy, smartlossy)

• Direct invocation of the FAPEC entropy coding core without any pre-processing:

  entropy_coder()

Examples

Compress and decompress a file

In this example we use the kmall option of FAPEC, suitable for this kind of geomaritime data files from Kongsberg Maritime:

from fapyc import Fapyc, Unfapyc, FapecLicense

filename = input("Path to KMALL file: ")

# Here we invoke FAPEC to directly run on files,
# so the memory usage will be small (just 16MB or so)
# although it won't allow us to directly access the
# (de)compressed buffers.
f = Fapyc(filename)
# Check that we have a valid license
lt = f.fapyc_get_lic_type()
if lt >= 0:
    ln = FapecLicense(lt).name
    lo = f.fapyc_get_lic_owner()
    print("FAPEC",ln,"license granted to",lo)
    f.compress_kmall()
    # Let's now decompress it, as a check
    print("Preparing to decompress %s" % (filename + ".fapec"))
    uf = Unfapyc(filename + ".fapec")
    uf.decompress(output=filename+".dec")
else:
    print("No valid license found")

Decompress an image into a buffer and show it

With this example we can view a colour image compressed with FAPEC:

from fapyc import Unfapyc
import numpy as np
from matplotlib import pyplot as plt

filename = input("Path to FAPEC-compressed 8-bit RGB image file: ")

# Decompress the file into a byte array buffer
uf = Unfapyc(filename = filename)

# Get the image features - assuming part index 0 (OK for a single-part archive; otherwise, we're simply taking the first part)
cmpOpts = uf.fapyc_get_part_cmpopts(0)

# Get the compression algorithm, which should be CILLIC, DWT or HPA for an image
algo = cmpOpts['algorithm'].decode('utf-8')
if algo != 'CILLIC' and algo != 'DWT' and algo != 'HPA':
    raise Exception("Not an image")
else:
    print("Found image compressed with the",algo,"algorithm")

# Get the image features we need
w = cmpOpts['imageWidth']
h = cmpOpts['imageHeight']
bpp = cmpOpts['sampleBits']
bands = cmpOpts['nBands']
coding = cmpOpts['bandsCoding']
coding2text = ['BIP','BIL','BSQ']

# Do some check
if bpp != 8 or bands != 3 or coding != 0:
    raise Exception("This test needs 8-bit colour images (3 colour bands) in pixel-interleaved coding mode")
else:
    print("Image features:",w,"x",h,"pixels,",bpp,"bits per pixel,",bands,"colour bands,",coding2text[coding],"coding")

uf.decompress()
# Check consistency (image dimensions vs. buffer size)
if len(uf.outputBuffer) != 3*w*h:
    print("Image dimensions inconsistent with file contents!")
else:
    # Reshape this one-dimensional array into a three-dimensional array (height, width, colours) to plot it
    ima = np.reshape(np.frombuffer(uf.outputBuffer, dtype=np.dtype('u1')), (h, w, 3))
    plt.imshow(ima)
    plt.show()

Compress and decompress a buffer

In this example we use the tab option of FAPEC, which typically outperforms gzip and bzip2 on tabulated text/numerical data such as point clouds or certain scientific data files:

from fapyc import Fapyc, Unfapyc

filename = input("Path to file: ")
file = open(filename, "rb")
# Beware - Load the whole file to memory
data = file.read()
f = Fapyc(buffer = data)
# Use 2 threads
f.fapyc_set_nthreads(2)
# Invoke our tabulated-text compression algorithm
# indicating a comma separator
f.compress_tabtxt(sep1=',')
print("Ratio =", round(float(len(data))/len(f.outputBuffer), 4))

# Now we decompress the buffer into another buffer
uf = Unfapyc(buffer = f.outputBuffer)
uf.fapyc_set_useropts(0, 3, 0, 0, 0)
uf.decompress()
print("Decompressed size:", len(uf.outputBuffer))

Decompress a file into a buffer, and do some operations on it

Here we provide a quite specific use case, based on the ESA/DPAC Gaia DR3 bulk catalogue (which is publicly available as FAPEC-compressed CSVs). In this example, we decompress two of the files, and while getting their CSV-formatted contents with Pandas we filter the contents according to some conditions, and generate some plots. This is just to illustrate how you can directly work on several compressed files. Note that it may require quite a lot of RAM, perhaps 4GB. You may need to install pyqt5 with pip.

from fapyc import Unfapyc
from io import BytesIO
import pandas as pd
import numpy as np
import matplotlib.pyplot as plt
from matplotlib import colors
import gc

filename = input("Path to GaiaDR3 csv.fapec file: ")
filename2 = input("Path to another GaiaDR3 csv.fapec file: ")

### Option 1: open the file, load it to memory (beware!), and decompress the buffer; it would be like this:
#file = open(filename, "rb")
#data = file.read()
#uf = Unfapyc(buffer = data)

### Option 2: directly decompress from the file into a buffer:
uf = Unfapyc(filename = filename)

# Here we'll use a verbose mode to see the decompression progress
uf.fapyc_set_useropts(2, 3, 0, 0, 0)
uf.fapyc_set_nthreads(2)
# Invoke decompressor
uf.decompress()

# Define our query (filter):
myq = "ra_error < 0.1 & dec_error < 0.1 & ruwe > 0.5 & ruwe < 2"

# Regenerate the CSV from the bytes buffer
print("Decoding and filtering CSV...")
df = pd.read_csv(BytesIO(uf.outputBuffer), comment="#").query(myq)

# Repeat for the 2nd file
uf = Unfapyc(filename = filename2)
uf.fapyc_set_useropts(2, 3, 0, 0, 0)
uf.fapyc_set_nthreads(2)
uf.decompress()
print("Decoding, filtering and joining CSV...")
df = pd.concat([df, pd.read_csv(BytesIO(uf.outputBuffer), comment="#").query(myq)])
# Remove NaNs and nulls from these two columns
df = df[np.isfinite(df['bp_rp'])]
df = df[np.isfinite(df['phot_g_mean_mag'])]
# Delete Unfapyc and force garbage collection, to try to free some memory
del uf
gc.collect()

print("Info from the filtered CSVs:")
print(df.info())

# Prepare some nice histograms for all data
plt.subplot(2,2,1)
plt.title("Skymap (%d sources)" % df.shape[0])
plt.xlabel("RA")
plt.ylabel("DEC")
print("Getting 2D histogram...")
plt.hist2d(df.ra, df.dec, bins=(200, 200), cmap=plt.cm.jet)
plt.colorbar()

plt.subplot(2,2,2)
plt.title("G-mag distribution")
plt.xlabel("G magnitude")
plt.ylabel("Counts")
plt.yscale("log")
print("Getting histogram...")
plt.hist(df.phot_g_mean_mag, bins=(100))

plt.subplot(2,2,3)
plt.title("Colour-Magnitude Diagram")
plt.xlabel("BP-RP")
plt.ylabel("G")
print("Getting 2D histogram...")
plt.hist2d(df.bp_rp, df.phot_g_mean_mag, bins=(100, 100), norm = colors.LogNorm(), cmap=plt.cm.jet)
plt.colorbar()

plt.subplot(2,2,4)
plt.title("Parallax error distribution")
plt.xlabel("G magnitude")
plt.ylabel("Parallax error")
print("Getting 2D histogram...")
plt.hist2d(df.phot_g_mean_mag, df.parallax_error, bins=(100, 100), norm = colors.LogNorm(), cmap=plt.cm.jet)

print("Plotting...")
plt.show()

Compress file using a logger

In this example, the user can provide a Python logger to get an information message from Fapyc, to capture the progress and get more information in case of errors (otherwise the native FAPEC library just writes to the console).

import logging
from fapyc import Fapyc, Unfapyc

filename = input("Path to the file to compress: ")
logger_file = 'fapyc.log'
logger = logging.getLogger(__name__)
logging.basicConfig(filename=logger_file, filemode='w', format='%(name)s - %(levelname)s - %(message)s')
logger.setLevel(logging.DEBUG)

file = open(filename, "rb")
data = file.read()
file.close()

f = Fapyc(filename = filename, logger = logger)
f.fapyc_set_loglev(logging.INFO)
f.compress_doubles(output = "a.fapec")