Sometimes you want to keep ongoing writing in a private repository
and make only selected chapters and/or files publicly visible.
In such cases
one can set up the book project structure in a private repository, but
use a public repository instead of the doc/pub
directory for publishing
selected compiled documents. This is easy: just change the dest=
line, where the publishing directory is defined, in all make*.sh
scripts in doc/src/chapters
. The files will then be copied to
this alternative destination.
Often, you want to publish the software associated with the book
project, stored in doc/src/chapter/nickname/src-nickname
, as a part
of the public repository. Such files can also easily be copied, say to
src/nickname
in the public repository. However, software files often
change names and locations in subdirectories, and then you need to be
very careful with updating the Git commands in the public repository
every time you do git add
or git rm
locally in the private
repository. This problem occurs with text files too, but maybe less
often, so the recipe given below applies to all kind of files you want
to mirror from a private to a public repository.
We have made a script rsync_git.py that can copy files from one repository to another and log files that are removed or deleted and then take the appropriate Git actions. Running
Terminal> rsync_git.py src-mychap $HOME/repos/pub/mybook/src/mychap
will copy all files from src-mychap
to $HOME/repos/pub/mybook/src/mychap
,
find which files that are new in src-mychap
and
must be added to the destination directory, and which files
that are removed in src-mychap
and should be removed in the destination
directory as well.
An rsync
command is run to the physical copy and removal of
files, followed by git add
and git rm
commands. In this way, you can automatically keep the public repository
as a mirror of parts of your private repository! [3]
3: This functionality should be a part of Git, but no Git expert I have talked to has ever seen use for merging a flexibly defined subset of a repository with another repository. (The current functionality of Git is not capable of working with, e.g., branches that merge with only parts of another branch.)
The rsync_git.py
script is listed below for reference. Note that
a file $HOME/.rsyncexclude
can be made to filter out certain files
that you never want to copy (this is always a good idea!).
#!/usr/bin/env python
"""
Sync two directory trees with rsync and perform corresponding
git operations (add or rm).
Skip files listed in $HOME/.rsyncexclude.
Usage: rsync_git.py from-dir to-dir
Example: rsync_git.py src-mychap $HOME/repos/pub/mybook/src/mychap
The from-dir is the source and the to-dir is the destination
(e.g. a public directory where resources are exposed).
The script must be run from a dir within the repo of to-dir.
"""
# Typical rsync output:
"""
sending incremental file list
deleting decay7.py
decay_TULL.py
sent 675 bytes received 34 bytes 1418.00 bytes/sec
total size is 94788 speedup is 133.69
"""
# Example on $HOME/.rsyncexclude file
"""
.#*
#*
*.rsync~
*.a
*.o
*.so
*~
.*~
*.log
*.dvi
*.aux
*.old
tmp_*
.tmp*
*.tar
*.tar.gz
*.tgz
*.pyc
"""
import commands, os, sys
from_ = sys.argv[1]
to_ = sys.argv[2]
cmd = 'rsync -rtDvz -u -e ssh -b ' + \
'--exclude-from=$HOME/.rsyncexclude ' + \
'--suffix=.rsync~ --delete --force %s/ %s' % (from_, to_)
print cmd
failure, output = commands.getstatusoutput(cmd)
print output
delete = []
add = []
for line in output.splitlines():
relevant_line = True
for text in 'sending incremental file list', \
'sent ', 'total size is':
if line.startswith(text):
relevant_line = False
if relevant_line and line != '':
if line.startswith('deleting'):
delete.append(line.split()[1])
else:
add.append(line.strip())
print delete
print add
for filename in delete:
option = '-rf' if os.path.isdir('%s/%s' % (to_, filename)) else '-f'
cmd = 'git rm %s %s/%s' % (option, to_, filename)
print cmd
os.system(cmd)
for filename in add:
cmd = 'git add %s/%s' % (to_, filename)
print cmd
os.system(cmd)