4

I regularly use a tool (Amesim) that packages its files in an uncompressed tar file. For versioning, I've typically named files as file1_Rev01.ame, and iterated up with changes. This works when I'm the only user, but lately I'm sharing files/models more regularly. Trying to share these models is painful, they often include results which are quite large (gbs of data), and keeping track of changes between versions if difficult unless rigorously adding text within the model on each change. (Amesim is a tool like Simulink.)

I've been reading up on git hooks and git filters, but I'm not sure what to do to better manage versioning of a tarball.

Let's say I have the file "my_file.tar" and it consists of a.txt, b.model, c.data, and d.results.

From the application side, I would stage "my_file.tar" and submit a commit "Updates to model." With no changes to git, this tracks the changes to a binary file. This isn't readable and consumes significant space. If results are included, the file is quite large. Cloning the repo will be challenging if results are continually stored.

For my first attempt, I tried to use pre-commit and post-checkout hooks.

At commit, my pre-commit hook untars "my_file.tar" into a directory "my_file_tar." It removes the *.results file which comes from running the model. It is unnecessary to track this and saves significant space (gbs).

When I pull the model, post-checkout will search for any folders with _tar and tar them, renaming them to my_file.tar.

Now generally this works. But, how should I handle my_file.tar and uncompressed folder? If I auto-delete the uncompressed folder after check-out, git states that I have significant changes to track. Do I need to add/remove the folder to .gitignore every time? Additionally, the tar file will never show that it is tracked, because I removed it in the pre-commit code. What can I do to clean up this process? How should I handle this differently?

References:

For this code, the .ame is a tar file.

pre-commit

#!/usr/bin/env python

import argparse
import os
import tarfile
import zipfile
import subprocess

def parse_args():
    pass

def log_file(log_item):
    cwd = os.getcwd()
    file = open("MyFile.txt", "a") # Open file in append mode
    file.write(log_item + '\n')
    return 1
    
def get_staged_ame_files():
    '''Request a list of staged files from git and return a list of *.ame files

    This function opens a subprocess with git, requests a list of names in the git staged list. It will return a list of strings.
    '''
    out = subprocess.Popen(['git', 'diff', '--staged', '--name-only'], stdout=subprocess.PIPE, stderr=subprocess.STDOUT)
    stdout, stderr = out.communicate()
    # Separate output by newlines
    # staged_files = stdout.split(b'\n') # split as bytes
    
    # filter for files with .ame 
    staged_files = stdout.decode('utf-8').split('\n') # split as strings
    # Create list of *just* amesim files
    staged_ame_files = []
    for entry in staged_files:
        if entry.endswith(".ame"):
            staged_ame_files.append(entry)
    
    if not staged_ame_files:
        return None
    else:
        return staged_ame_files

def extract_ame_files(file_list):
    folder_list = []
    for list_item in file_list:
        # If file exists, extract it. Else continue.
        if os.path.isfile(list_item):
            tar = tarfile.open(list_item, "r:")
            folder_name = list_item[0:-4] + "_ame"
            folder_list.append(folder_name)
            tar.extractall(path = folder_name)
            tar.close()
            log_file(folder_name)
        else:
            print("File {} does not exist.".format(list_item))
            
    return folder_list
    

def cleanup_ame_ignored_files(folder_list):
    '''Removes unecessary files from the folder. 
    
    '''
    for folder in folder_list:
        file_list = os.listdir(folder)
        for file in file_list:
            if item.endswith(".results"):
                os.remove(item)
            if item.endswith(".exe"):
                os.remove(item)
    return 1


def git_add_ame_folders(folders):
    # Add *_ame folders to git stage
    for folder in folders:
        out = subprocess.Popen(['git', 'add', folder + '/'], stdout=subprocess.PIPE, stderr=subprocess.STDOUT)
        stdout, stderr = out.communicate()
        # The -u will capture removed files?
        out = subprocess.Popen(['git', 'add', '-u', folder + '/'], stdout=subprocess.PIPE, stderr=subprocess.STDOUT)
        stdout, stderr = out.communicate()
        
        log_file(stdout.decode('utf-8'))
    return 1
    
def remove_ame_from_staging(file_list):
    # Loop through any staged ame files.
    for file in file_list:
        out = subprocess.Popen(['git', 'rm', '--cached', file], stdout=subprocess.PIPE, stderr=subprocess.STDOUT)
        stdout, stderr = out.communicate()
    return 1

def main(args=None):
    # if file name is *.ame
    # extract *.ame as a tar of the same name into a folder of the same name + _ame
    # delete .results file
    # don't commit .ame file 
    
    # Search for files we want to process in the staged list
    # These will only be *.ame files.
    staged_ame_files = get_staged_ame_files()
    if not staged_ame_files:
        # If its empty, there's nothing to do. End the function.
        return 0
    
    # We're not empty, lets extract each one.
    folder_list = extract_ame_files(staged_ame_files)
    
    # Delete all .results files in each extracted folder  path
    
    # Stage all files in each folder path 
    git_add_ame_folders(folder_list)
    
    # Unstage the .ame file
    remove_ame_from_staging(staged_ame_files)
    return 1

if __name__ == "__main__":
    args = parse_args()
    main(args)

and post-checkout

#!/usr/bin/env python

import argparse
import os
import tarfile
import zipfile
import subprocess
import shutil
#from shutil import rmtree # Delete directory trees

def parse_args():
    pass

def log_file(log_item):
    cwd = os.getcwd()
    file = open("MyFile2.txt", "a") # Open file in append mode
    file.write(log_item + '\n')
    return 1
    
def compress_ame_files(folder_list):
    for list_item in folder_list:
        log_file("We're on item {}".format(list_item))
        file_name = list_item[0:-4] + ".ame"
        log_file("Tar file name {}".format(file_name))
        # Delete the file if it exists first.
        os.remove(file_name)
        with tarfile.open(file_name, "w:") as tar:
            tar.add(list_item, arcname=os.path.basename('../'))
    return 1
    

def cleanup_ame_ignored_files(folder_list):
    '''Removes unecessary files from the folder. 
    
    '''
    for folder in folder_list:
        file_list = os.listdir(folder)
        for file in file_list:
            if item.endswith(".results"):
                os.remove(item)
            if item.endswith(".exe"):
                os.remove(item)
    return 1


def git_add_ame_folders(folders):
    # Add *_ame folders to git stage
    for folder in folders:
        out = subprocess.Popen(['git', 'add', folder + '/'], stdout=subprocess.PIPE, stderr=subprocess.STDOUT)
        stdout, stderr = out.communicate()
        # The -u will capture removed files?
        out = subprocess.Popen(['git', 'add', '-u', folder + '/'], stdout=subprocess.PIPE, stderr=subprocess.STDOUT)
        stdout, stderr = out.communicate()
        
        #log_file(stdout.decode('utf-8'))
    return 1
    
def remove_ame_from_staging(file_list):
    # Loop through any staged ame files.
    for file in file_list:
        out = subprocess.Popen(['git', 'rm', '--cached', file], stdout=subprocess.PIPE, stderr=subprocess.STDOUT)
        stdout, stderr = out.communicate()
    return 1

def fast_scandir(dirname):
    # https://stackoverflow.com/questions/973473/getting-a-list-of-all-subdirectories-in-the-current-directory?rq=1
    subfolders= [f.path for f in os.scandir(dirname) if f.is_dir()]
    for dirname in list(subfolders):
        subfolders.extend(fast_scandir(dirname))
    return subfolders

def delete_ame_folders(folders):
    for folder in folders:
        try:
            shutil.rmtree(folder)
        except OSError as e:
            print("Error: %s : %s" % (dir_path, e.strerror))
    return 1
    
#def main(args=None):
def main(lines):
    print("Post checkout running.")
    # find folders with the name _ame
    #log_file("We're running.")
    folder_list = []
    for folder in fast_scandir(os.getcwd()):
        if folder.endswith("_ame"):
            #log_file("Found folder {}.".format(folder))
            folder_list.append(os.path.join(os.getcwd(), folder))
    # tar each folder up and rename with .ame
    compress_ame_files(folder_list)
    
    # Delete the folders
    #delete_ame_folders(folder_list)

    return 1

if __name__ == "__main__":
    args = parse_args()
    main(args)
Rukie
  • 91
  • 7
  • It seems the Zippey method of using a git filter is better than using git hooks. I should be able to use tar to remove the .results files without extracting it... more to come. As an additional bonus, the .gitattributes can be tracked within the git directory and more easily shared with other teammates. It seems. – Rukie Dec 23 '20 at 05:10

1 Answers1

3

The code in this answer implements a git filter as opposed to a pre-commit hook and a post-checkout hook in the question. The advantage of the filter is that it only manipulates one file. Additional files do not need to be tracked and committed/pulled separately. Rather, like Zippey, it creates an uncompressed data stream and removes unnecessary files along the way.

Note: Don't use print statements, as it messes with the stdout stream in the git filter. This was a painful lesson.

Note: CRLF and LF endings are a problem. When decoding from the first git pull, I had to clean the line endings because Sourcetree/Git converted to windows format.

Discussion of solution:

Since the file I am working with is an uncompressed tar, the Zippey solution didn't apply directly. Zippey is only for zip files. I implemented Zippey's technique with tar files instead.

On commit, a clean filer is applied that 'encodes' the tar file. The encode function takes each file and records the length of the data, the raw length of the data if binary, the mode of storing (ascii or binary), and the filename.

The encode script streams all files into a single file of the same name in an uncompressed format. Binary files are base64 encoded into a single line, making diffs easier to read.

During encode, files of specific extensions are avoided (like results files).

On pull, a smudge filter decompresses the file by utilizing the four meta tags to read through the information. Each file is processed and added to a tar file object, and at the end a tar file is written out.

Like Zippey, on a fresh clone of the repository an encoded file is pulled that's unreadable to my tool. So Clone Setup looks for my *.ame files that are encoded and decodes them, as well as sets up appropriate git filters.

As I work on both linux and windows machines, and git has a tendency to add CRLF on checkout, the scripts make sure to remove CRLF before encoding, and removing CRLF from encoded files before decoding.

amefilter.py

import tarfile
import sys
import io
import base64
import string
import tempfile
import os.path

DEBUG_AME_FILTER = False
NAME = 'Amesim_Git'
ENCODING = 'UTF-8'

W_EOL = b'\r\n'
U_EOL = b'\n'

# decompress these defined files
AME_EXTENSIONS = ['.amegp', '.cir', '.sad', '.units', '.views', '.xml']
ASCII_EXTENSIONS = ['.txt', '.py']
# Do not include these files in tracking. 
EXCLUDE = ['.results']

def debug(msg):
    '''Print debug message'''
    if DEBUG_AME_FILTER:
        sys.stderr.write('{0}: debug: {1}\n'.format(NAME, msg))

def error(msg):
    '''Print error message'''
    sys.stderr.write('{0}: error: {1}\n'.format(NAME, msg))

def init():
    '''Initialize writing; set binary mode for windows'''
    debug("Running on {}".format(sys.platform))
    if sys.platform.startswith('win'):
        import msvcrt
        debug("Enable Windows binary workaround")
        msvcrt.setmode(sys.stdin.fileno(), os.O_BINARY)
        msvcrt.setmode(sys.stdout.fileno(), os.O_BINARY)

def encode(input, output):
    '''Encode into special VCS friendly format from input to output'''
    debug("ENCODE was called")
    # Create a temporary file based off of the input AME file
    # This lets tarfile access a binary file object
    tfp = tempfile.TemporaryFile(mode='w+b')
    # Write contents into temporary file
    tfp.write(input.read())
    tfp.seek(0)  # Make sure tarfile reads from the start, otherwise object is empty
    tar = tarfile.open(fileobj=tfp, mode='r:')
    # Loop through objects within tar file
    for name in tar.getnames():
        # Get the file name of each object.
        tarinfo = tar.getmember(name)
        if tarinfo.isdir():
            continue # Skip folders, not sure how to handle encode/decode yet.
        data = tar.extractfile(name).read()
        
        # List of ASCII files to decode and version control
        text_extensions = list(set(AME_EXTENSIONS).union(set(ASCII_EXTENSIONS)))
        
        # Isolate extension.
        extension = os.path.splitext(name)[1][1:].strip().lower()
        # Amesim may store batched simulations as *.results.1, *.results.2, remove numeric endings and identify the real ending.
        if extension.isnumeric():
            root_name = os.path.splitext(name)[0][0:]
            real_extension = os.path.splitext(root_name)[1][1:].strip().lower()
            if real_extension in EXCLUDE:
                continue  # Skip excluded extensions
            
        if extension in EXCLUDE:
            continue  # Skip excluded extensions.
            
        # Encode the defined extensions in UTF-8
        try:
            # Check if text data
            data.decode(ENCODING)
            data = data.replace(W_EOL, U_EOL)  # Fix line endings
            try:
                strdata = map(chr, data)
            except TypeError:
                strdata = data
            if extension not in text_extensions and not all(c in string.printable for c in strdata):
                # File is not ascii, append binary file.
                raise UnicodeDecodeError(ENCODING, "".encode(ENCODING), 0, 1, "Artificial exception")

            # Encode
            debug("Appending text file '{}'".format(name))
            mode = 'A'  # ASCII Mode
            output.write("{}|{}|{}|{}\n".format(len(data), len(data), mode, name).encode(ENCODING))
            output.write(data)
            output.write("\n".encode(ENCODING)) # Separation from next meta line
        except UnicodeDecodeError:
            # Binary data
            debug("Appending binary file '{}'".format(name))
            mode = 'B'  # Binary Mode
            raw_len = len(data)
            data = base64.b64encode(data)
            output.write("{}|{}|{}|{}\n".format(len(data), raw_len, mode, name).encode(ENCODING))
            output.write(data)
            output.write("\n".encode(ENCODING))  # Separation from next meta line
    tar.close()

def decode(input, output):
    '''Decode from special VCS friendly format from input to output'''
    debug("DECODE was called")
    tfp = tempfile.TemporaryFile(mode='w+b')
    tar = tarfile.open(fileobj=tfp, mode='w:')
    #input = io.open(input, 'rb')
    while True:
        meta = input.readline().decode(ENCODING)
        if not meta:
            break
        #print(meta)
        (data_len, raw_len, mode, name) = [t(s) for (t, s) in zip((int, int, str, str), meta.split('|'))]
        #print('Data length:{}'.format(data_len))
        #print('Mode: {}'.format(mode))
        #print('Name: {}'.format(name))
        if mode == 'A':
            #print('Appending ascii data')
            debug("Appending text file '{}'".format(name))
            #https://stackoverflow.com/questions/740820/python-write-string-directly-to-tarfile
            info = tarfile.TarInfo(name=name.rstrip())
            info.size = raw_len
            raw_data = input.read(data_len)
            binary_data = io.BytesIO(raw_data)
            # Add each file object to our tarball
            tar.addfile(tarinfo=info, fileobj=binary_data)
            input.read(1) # Skip last '\n'
        elif mode == 'B':
            #print('Appending binary data')
            debug("Appending binary file '{}'".format(name.rstrip()))

            info = tarfile.TarInfo(name=name.rstrip())
            info.size = raw_len
            raw_data = input.read(data_len)
            decoded_data = base64.b64decode(raw_data)
            binary_data = io.BytesIO(decoded_data)
            tar.addfile(tarinfo=info, fileobj=binary_data)
            input.read(1) # Skip last '\n'
        else:
            # Should never reach here
            tar.close()
            tfp.close()
            error('Illegal mode "{}"'.format(mode))
            sys.exit(1)

    # Flush all writes
    tar.close()

    # Write output
    tfp.seek(0) # Go to the start of our temporary file
    output.write(tfp.read())
    tfp.close()

def main():
    '''Main program'''
    #import codecs
    #sys.stdout = codecs.getwriter('utf8')(sys.stdout)
    init()
    input = io.open(sys.stdin.fileno(), 'rb')
    output = io.open(sys.stdout.fileno(), 'wb')
    if len(sys.argv) < 2 or sys.argv[1] == '-' or sys.argv[1] == '--help':
        # This is wrong
        sys.stdout.write("{}\nTo encode: 'python ame_filter.py e'\nTo decode: 'python ame_filter.py d'\nAll files read from stdin and printed to stdout\n".format(NAME))
    elif sys.argv[1] == 'e':
        encode(input, output)
    elif sys.argv[1] == 'd':
        decode(input, output)
    else:
        error("Illegal argument '{}'. Try --help for more information".format(sys.argv[1]))
        sys.exit(1)

        
if __name__ == '__main__':
    main()

Clone_Setup.py

#!/usr/bin/env python
'''
Clone_Setup.py initializes the git environment. 
Each time a new instance of the repository is generated, these commands must 
be run.

'''
import os
import sys
import io
import subprocess
import ame_filter as amef
import tempfile
import shutil

# replacement strings
W_EOL = b'\r\n'
U_EOL = b'\n'

def setup_git():
    os.system("git config filter.ame_filter.smudge \"./ame_filter.py d\"")
    os.system("git config filter.ame_filter.clean \"./ame_filter.py e\"")
    
    '''
        Create .gitattributes programmatically. 
        Add these lines if they do not exist
    '''
    items = ["*.ame filter=ame_filter", "*.ame diff"]
    try:
        with open(".gitattributes", "x") as f:
            for item in items:
                f.write(item + "\n")
    except:
        with open(".gitattributes", "r+") as f:
            for item in items:
                f.seek(0)
                line_found = any(item in line for line in f)
                if not line_found:
                    f.seek(0, os.SEEK_END)
                    f.write("\n" + item)
    
    '''
        Create .gitignore programmatically. 
        Add these lines if they do not exist.
    '''
    items = ["*.gra",
             "*.res",
             "*.req",
             "*.pyc",
             "*.results",
             "*.results.*"
             ]
    
    try:
        with open(".gitignore", "x") as f:
            for item in items:
                f.write(item + "\n")
    except:
        with open(".gitignore", "r+") as f:
            for item in items:
                f.seek(0)
                line_found = any(item in line for line in f)
                if not line_found:
                    f.seek(0, os.SEEK_END)
                    f.write("\n" + item)
        

''' Search for AME files '''
def find_ame_files():
    out = subprocess.Popen(['git', 'ls-files'], stdout=subprocess.PIPE, stderr=subprocess.STDOUT)
    stdout, stderr = out.communicate()
    # Separate output by newlines
    # filter for files with .ame 
    git_files = stdout.decode('utf-8').split('\n') # split as strings
    # Create list of *just* amesim files
    ame_files = [entry for entry in git_files if entry.endswith(".ame")]
    ''' #  Equivalent code block
    ame_files = []
    for entry in git_files:
        if entry.endswith(".ame"):
            ame_files.append(entry)
    '''
    return ame_files

def decode_ame_files(ame_files):
    for file in ame_files:
        input = io.open(file, 'rb')
        tfp = tempfile.TemporaryFile(mode='w+b')
        # Write contents into temporary file
        tfp.write(input.read().replace(W_EOL, U_EOL))
        tfp.seek(0)
        input.close()
        output = io.open(file+'~', 'wb')
        try:
            amef.decode(tfp, output)
            output.close()
            shutil.move(file+'~', file)
        except:
            print("File is already decoded. Returning to normal.")
            output.close()
        finally:
            os.remove(file+'~')
            
            

def main():
    '''Main program'''
    print("Setting up git.")
    setup_git()
    print("Finding ame files.")
    ame_files = find_ame_files()
    print(ame_files)
    print("Decoding ame files.")
    decode_ame_files(ame_files)
    
        
if __name__ == '__main__':
    main()
    # Keep console open to view messages on windows machines.
    if os.name == 'nt':
        input("Press enter to exit")
Rukie
  • 91
  • 7
  • This is an interesting problem from what I've read. I think your answer can be improved if it includes some more words to explain what you did. – Code-Apprentice Dec 30 '20 at 23:32
  • @Code-Apprentice I've added some commentary to the answer. Is this what you're looking for? – Rukie Dec 31 '20 at 01:24
  • The background information probably belongs in the question rather than the answer. As for commentary, I had more in mind commenting the code in the answer to explain how it solves the problem. Is the code in the answer different from that in the question? If so, words to highlight the differences will be helpful for future visitors. If not, then why is the code posted twice? Should you remove it from the question and only post it in the answer instead? – Code-Apprentice Dec 31 '20 at 02:33
  • Ah, okay. The code in the answer implements a git filter as opposed to a pre-commit hook and a post-checkout hook in the question. The implementation of the filter is cleaner and results are more easily viewed within a git diff. I'll take some time in the morning to fix this up. – Rukie Dec 31 '20 at 02:48
  • Yah, that info directly in the answer will make it more useful to future readers – Code-Apprentice Dec 31 '20 at 04:45