def addLogEntry(self, entry, save=True):
"""This is called when a new log entry is created"""
# create log directory if it doesn't exist
# error will be thrown if this is not possible
_busy = Purr.BusyIndicator()
self._initIndexDir()
# discard temporary watchers -- these are only used to keep track of
# deleted files
self.temp_watchers = {}
# ignored entries are only there to carry info on ignored data products
# All we do is save them, and update DP policies based on them
if entry.ignore:
entry.save(self.logdir)
# proper entries are added to list
else:
self.entries.append(entry)
Purr.progressMessage("Saving new log entry")
# find previous entry -- skip over "ignore" entries
for prev in self.entries[-2::-1]:
if not prev.ignore:
break
else:
prev = None
entry.setLogDirectory(self.logdir)
entry.setPrevUpNextLinks(prev=prev, up=os.path.join("..", Purr.RenderIndex.INDEX))
entry.save()
self.timestamp = self.last_scan_timestamp
# regenerate links of previous entry
prev and prev.generateIndex()
# and our log may need to be regenerated
if save:
self.save()
self.updatePoliciesFromEntry(entry, new=True)
def addLogEntry(self, entry, save=True):
"""This is called when a new log entry is created"""
# create log directory if it doesn't exist
# error will be thrown if this is not possible
_busy = Purr.BusyIndicator()
self._initIndexDir()
# discard temporary watchers -- these are only used to keep track of
# deleted files
self.temp_watchers = {}
# ignored entries are only there to carry info on ignored data products
# All we do is save them, and update DP policies based on them
if entry.ignore:
entry.save(self.logdir)
# proper entries are added to list
else:
self.entries.append(entry)
Purr.progressMessage("Saving new log entry")
# find previous entry -- skip over "ignore" entries
for prev in self.entries[-2::-1]:
if not prev.ignore:
break
else:
prev = None
entry.setLogDirectory(self.logdir)
entry.setPrevUpNextLinks(prev=prev,
up=os.path.join("..",
Purr.RenderIndex.INDEX))
entry.save()
self.timestamp = self.last_scan_timestamp
# regenerate links of previous entry
prev and prev.generateIndex()
# and our log may need to be regenerated
if save:
self.save()
self.updatePoliciesFromEntry(entry, new=True)
def makeDataProducts(self, files, unbanish=False, unignore=False):
"""makes a list of DPs from a list of (filename,quiet) pairs.
If unbanish is False, DPs with a default "banish" policy will be skipped.
Symlinks will be resolved, and non-unique filenames removed from list.
"""
paths = set()
dps = []
for filename, quiet in files:
filename = filename.rstrip('/')
sourcepath = Purr.canonizePath(filename)
if sourcepath not in paths:
paths.add(sourcepath)
filename = os.path.basename(filename)
policy, filename, comment = self._default_dp_props.get(
filename, ("copy", filename, ""))
dprintf(4, "%s: default policy is %s,%s,%s\n", sourcepath,
policy, filename, comment)
if policy == "banish":
if unbanish:
policy = "copy"
else:
continue
if unignore and policy == "ignore":
policy = "copy"
dps.append(
Purr.DataProduct(filename=filename,
sourcepath=sourcepath,
policy=policy,
comment=comment,
quiet=quiet))
return sorted(dps, lambda a, b: cmp(a.filename, b.filename))
def addWatchedDirectory(self, dirname, watching=Purr.WATCHED, save_config=True):
"""Starts watching the specified directories for changes"""
# see if we're alredy watching this exact set of directories -- do nothing if so
dirname = Purr.canonizePath(dirname)
# do nothing if already watching
if dirname in self.watched_dirs:
dprint(1, "addWatchDirectory(): already watching %s\n", dirname)
# watching=None means do not change the watch-state
if watching is None:
return
else:
if watching is None:
watching = Purr.WATCHED
# make watcher object
wdir = Purrer.WatchedDir(dirname, mtime=self.timestamp,
watch_patterns=self._watch_patterns, ignore_patterns=self._ignore_patterns)
# fileset=None indicates error reading directory, so ignore it
if wdir.fileset is None:
print("There was an error reading the directory %s, will stop watching it." % dirname)
self.setWatchingState(dirname, Purr.REMOVED, save_config=True)
return
self.watchers[dirname] = wdir
self.watched_dirs.append(dirname)
dprintf(2, "watching directory %s, mtime %s, %d files\n",
dirname, time.strftime("%x %X", time.localtime(wdir.mtime)), len(wdir.fileset))
# find files in this directory matching the watch_patterns, and watch them for changes
watchset = set()
for patt in self._watch_patterns:
watchset.update(fnmatch.filter(wdir.fileset, patt))
for fname in watchset:
quiet = matches_patterns(fname, self._quiet_patterns)
fullname = Purr.canonizePath(os.path.join(dirname, fname))
if fullname not in self.watchers:
wfile = Purrer.WatchedFile(fullname, quiet=quiet, mtime=self.timestamp)
self.watchers[fullname] = wfile
dprintf(3, "watching file %s, timestamp %s, quiet %d\n",
fullname, time.strftime("%x %X", time.localtime(wfile.mtime)), quiet)
# find subdirectories matching the subdir_patterns, and watch them for changes
for fname in wdir.fileset:
fullname = Purr.canonizePath(os.path.join(dirname, fname))
if os.path.isdir(fullname):
for desc, dir_patts, canary_patts in self._subdir_patterns:
if matches_patterns(fname, dir_patts):
quiet = matches_patterns(fname, self._quiet_patterns)
wdir = Purrer.WatchedSubdir(fullname, canary_patterns=canary_patts, quiet=quiet,
mtime=self.timestamp)
self.watchers[fullname] = wdir
dprintf(3, "watching subdirectory %s/{%s}, timestamp %s, quiet %d\n",
fullname, ",".join(canary_patts),
time.strftime("%x %X", time.localtime(wdir.mtime)), quiet)
break
# set state and save config
self.setWatchingState(dirname, watching, save_config=save_config)
def regenerate (self):
Purr.progressMessage("reading %s"%self.dp.filename,sub=True);
# init fitsfile to None, so that _read() above is forced to re-read it
fitsfile = pyfits.open(self.dp.fullpath);
header = fitsfile[0].header;
dprintf(3,"beginning render of",self.dp.fullpath); t0 = time.time();
# write out FITS header
self.headerfile,path,uptodate = self.subproduct("-fitsheader.html");
if not uptodate:
title = "FITS header for %s"%self.dp.filename;
html = """<HTML><BODY><TITLE>%s</TITLE>
<H2>%s</H2>
<PRE>"""%(title,title);
for line in header.ascard:
line = str(line).replace("<","<").replace(">",">");
html += line+"\n";
html += """
</PRE></BODY></HTML>\n""";
try:
file(path,"w").write(html);
except:
print "Error writing file %s"%path;
traceback.print_exc();
self.headerfile = None;
# figure out number of images to include
ndim = header['NAXIS'];
fitsshape = [ header['NAXIS%d'%i] for i in range(1,ndim+1) ];
self.cubesize = 'x'.join(map(str,fitsshape));
if ndim < 2:
raise TypeError,"can't render one-dimensional FITS files""";
elif ndim == 2:
fitsdata_to_images = lambda fdata:[fdata];
nplanes = 1;
else:
ax1 = ax2 = None;
# find the X/Y axes, by looking at CTYPEx
# note that the array axes are in reverse order. I.e. if X is FITS axis 1 and Y is axis 2,
# the array will be of e.g. shape 1,1,NY,NX, while fitsshape is [NX,NY,1,1]
for i in range(1,ndim+1):
ctype = header['CTYPE%d'%i];
if [ prefix for prefix in "RA","GLON","ELON","HLON","SLON" if ctype.startswith(prefix) ] \
or ctype in ("L","X"):
ax1 = ndim-i;
elif [ prefix for prefix in "DEC","GLAT","ELAT","HLAT","SLAT" if ctype.startswith(prefix) ] \
or ctype in ("M","Y"):
ax2 = ndim-i;
def __init__(self, filename=None, sourcepath=None, fullpath=None,
policy="copy", comment="",
timestamp=None, render=None,
quiet=False, archived=False):
# This is the absolute pathname to the original data product
self.sourcepath = Purr.canonizePath(sourcepath)
# Base filename (w/o path) of data product within the log storage area.
# Products may be renamed when they are moved or copied over to the log.
self.filename = filename or (sourcepath and os.path.basename(sourcepath))
# Full path to the DP within the log storage area.
# This is None until a DP has been saved.
self.fullpath = fullpath
# Handling policy for DP: "copy","move","ignore", etc.
self.policy = policy
# Comment associated with DP
self.comment = comment
# Once a DP has been saved, this is the timestamp of data product at time of copy
self.timestamp = timestamp
# Name of renderer used to render this DP.
self.render = render
# if True, DP is watched quietly (i.e. Purr does not pop up windows on update)
self.quiet = quiet
# if True, DP has already been archived. This is False for new DPs until they're saved.
self.archived = archived
# if True, dp is ignored (policy is "ignore" or "banish")
# not that policy should not be changed after a DP has been created
self.ignored = policy in ("ignore", "banish")
def _handle_start_DP(self,
filename=None,
src=None,
policy=None,
quiet=False,
timestamp=0,
comment=None,
render=None,
**kw):
# dispence with previous DP tag, if any
self._add_data_product()
# setup data for this tag
comment = comment or ""
try:
timestamp = float(timestamp)
except:
timestamp = time.time()
if not isinstance(quiet, bool):
try:
quiet = bool(int(quiet))
except:
quiet = bool(quiet)
comment = comment.replace("<", "<").replace(">", ">")
self._new_dp = Purr.DataProduct(filename=filename,
sourcepath=src,
timestamp=timestamp,
comment=comment,
fullpath=os.path.join(
self._dirname, filename or ""),
policy=policy,
render=render,
quiet=quiet,
archived=True)
def save(self, refresh=False):
"""Saves the log.
If refresh is set to a timestamp, will regenerate everything from scratch.
"""
# create directory if it doesn't exist
# error will be thrown if this is not possible
_busy = Purr.BusyIndicator()
Purr.progressMessage("Generating index in %s" % self.logdir)
self._initIndexDir()
# if refresh is True, re-save all entries.
if refresh:
refresh = time.time()
for i, entry in enumerate(self.entries):
entry.save(refresh=refresh)
Purr.RenderIndex.writeLogIndex(self.logdir, self.logtitle, self.timestamp, self.entries, refresh=refresh)
Purr.progressMessage("Wrote %s" % self.logdir)
def makeDataProducts(self, files, unbanish=False, unignore=False):
"""makes a list of DPs from a list of (filename,quiet) pairs.
If unbanish is False, DPs with a default "banish" policy will be skipped.
Symlinks will be resolved, and non-unique filenames removed from list.
"""
paths = set()
dps = []
for filename, quiet in files:
filename = filename.rstrip('/')
sourcepath = Purr.canonizePath(filename)
if sourcepath not in paths:
paths.add(sourcepath)
filename = os.path.basename(filename)
policy, filename, comment = self._default_dp_props.get(filename, ("copy", filename, ""))
dprintf(4, "%s: default policy is %s,%s,%s\n", sourcepath, policy, filename, comment)
if policy == "banish":
if unbanish:
policy = "copy"
else:
continue
if unignore and policy == "ignore":
policy = "copy"
dps.append(Purr.DataProduct(filename=filename, sourcepath=sourcepath,
policy=policy, comment=comment, quiet=quiet))
return sorted(dps, lambda a, b: cmp(a.filename, b.filename))
def __init__(self,
filename=None,
sourcepath=None,
fullpath=None,
policy="copy",
comment="",
timestamp=None,
render=None,
quiet=False,
archived=False):
# This is the absolute pathname to the original data product
self.sourcepath = Purr.canonizePath(sourcepath)
# Base filename (w/o path) of data product within the log storage area.
# Products may be renamed when they are moved or copied over to the log.
self.filename = filename or (sourcepath
and os.path.basename(sourcepath))
# Full path to the DP within the log storage area.
# This is None until a DP has been saved.
self.fullpath = fullpath
# Handling policy for DP: "copy","move","ignore", etc.
self.policy = policy
# Comment associated with DP
self.comment = comment
# Once a DP has been saved, this is the timestamp of data product at time of copy
self.timestamp = timestamp
# Name of renderer used to render this DP.
self.render = render
# if True, DP is watched quietly (i.e. Purr does not pop up windows on update)
self.quiet = quiet
# if True, DP has already been archived. This is False for new DPs until they're saved.
self.archived = archived
# if True, dp is ignored (policy is "ignore" or "banish")
# not that policy should not be changed after a DP has been created
self.ignored = policy in ("ignore", "banish")
def save(self, refresh=False):
"""Saves the log.
If refresh is set to a timestamp, will regenerate everything from scratch.
"""
# create directory if it doesn't exist
# error will be thrown if this is not possible
_busy = Purr.BusyIndicator()
Purr.progressMessage("Generating index in %s" % self.logdir)
self._initIndexDir()
# if refresh is True, re-save all entries.
if refresh:
refresh = time.time()
for i, entry in enumerate(self.entries):
entry.save(refresh=refresh)
Purr.RenderIndex.writeLogIndex(self.logdir,
self.logtitle,
self.timestamp,
self.entries,
refresh=refresh)
Purr.progressMessage("Wrote %s" % self.logdir)
def attachPurrlog(self, purrlog, watchdirs=[]):
"""Attaches Purr to the given purrlog directory. Arguments are passed to Purrer object as is."""
# check purrer stack for a Purrer already watching this directory
dprint(1, "attaching to purrlog", purrlog)
for i, purrer in enumerate(self.purrer_stack):
if os.path.samefile(purrer.logdir, purrlog):
dprint(1, "Purrer object found on stack (#%d),reusing\n", i)
# found? move to front of stack
self.purrer_stack.pop(i)
self.purrer_stack.insert(0, purrer)
# update purrer with watched directories, in case they have changed
for dd in (watchdirs or []):
purrer.addWatchedDirectory(dd, watching=None)
break
# no purrer found, make a new one
else:
dprint(1, "creating new Purrer object")
try:
purrer = Purr.Purrer(purrlog, watchdirs)
except Purr.Purrer.LockedError as err:
# check that we could attach, display message if not
QMessageBox.warning(
self, "Catfight!",
"""<P><NOBR>It appears that another PURR process (%s)</NOBR>
is already attached to <tt>%s</tt>, so we're not allowed to touch it. You should exit the other PURR
process first.</P>""" % (err.args[0], os.path.abspath(purrlog)),
QMessageBox.Ok, 0)
return False
except Purr.Purrer.LockFailError as err:
QMessageBox.warning(
self, "Failed to obtain lock",
"""<P><NOBR>PURR was unable to obtain a lock</NOBR>
on directory <tt>%s</tt> (error was "%s"). The most likely cause is insufficient permissions.</P>"""
% (os.path.abspath(purrlog), err.args[0]), QMessageBox.Ok,
0)
return False
self.purrer_stack.insert(0, purrer)
# discard end of stack
self.purrer_stack = self.purrer_stack[:3]
# attach signals
self.connect(purrer, SIGNAL("disappearedFile"),
self.new_entry_dialog.dropDataProducts)
self.connect(purrer, SIGNAL("disappearedFile"),
self.view_entry_dialog.dropDataProducts)
# have we changed the current purrer? Update our state then
# reopen Purr pipes
self.purrpipes = {}
for dd, state in purrer.watchedDirectories():
self.purrpipes[dd] = Purr.Pipe.open(dd)
if purrer is not self.purrer:
self.message("Attached to %s" % purrer.logdir, ms=10000)
dprint(1, "current Purrer changed, updating state")
# set window title
path = Kittens.utils.collapseuser(os.path.join(purrer.logdir, ''))
self.setWindowTitle("PURR - %s" % path)
# other init
self.purrer = purrer
self.new_entry_dialog.hide()
self.new_entry_dialog.reset()
dirs = [path for path, state in purrer.watchedDirectories()]
self.new_entry_dialog.setDefaultDirs(*dirs)
self.view_entry_dialog.setDefaultDirs(*dirs)
self.view_entry_dialog.hide()
self.viewer_dialog.hide()
self._viewing_ientry = None
self._setEntries(self.purrer.getLogEntries())
# print self._index_paths
self._viewer_timestamp = None
self._updateViewer()
self._updateNames()
# update directory widgets
self.wdirlist.clear()
for pathname, state in purrer.watchedDirectories():
self.wdirlist.add(pathname, state)
# Reset _pounce to false -- this will cause checkPounceStatus() into a rescan
self._pounce = False
self._checkPounceStatus()
return True
def save(self, dirname=None, refresh=0, refresh_index=True, emit_message=True):
"""Saves entry in the given directory. Data products will be copied over if not
residing in that directory.
'refresh' is a timestamp, passed to renderIndex(), causing all data products OLDER than the specified time to be regenerated.
'refresh_index', if true, causes index files to be re-rendered unconditionally
"""
if not refresh and not self.updated:
return
timestr = time.strftime("%Y%m%d-%H%M%S", time.localtime(self.timestamp))
Purr.progressMessage("Rendering entry for %s" % timestr)
if dirname:
self.pathname = pathname = os.path.join(dirname, "%s-%s" %
(("ignore" if self.ignore else "entry"), timestr))
elif not self.pathname:
raise ValueError("Cannot save entry: pathname not specified")
else:
pathname = self.pathname
# set timestamp
if not self.timestamp:
self.timestamp = int(time.time())
# get canonized path to output directory
pathname = Purr.canonizePath(pathname)
if not os.path.exists(pathname):
os.mkdir(pathname)
# now save content
# get device of pathname -- need to know whether we move or copy
devnum = os.stat(pathname).st_dev
# copy data products as needed
dprintf(2, "saving entry %s, %d data products\n", pathname, len(self.dps))
dps = []
for dp in self.dps:
# if archived, this indicates a previously saved data product, so ignore it
# if ignored, no need to save the DP -- but keep it in list
if dp.archived or dp.ignored:
dprintf(3, "dp %s is archived or ignored, skipping\n", dp.sourcepath)
dps.append(dp)
continue
# file missing for some reason (perhaps it got removed on us?) skip data product entirely
if not os.path.exists(dp.sourcepath):
dprintf(2, "data product %s missing, ignoring\n", dp.sourcepath)
continue
Purr.progressMessage("archiving %s" % dp.filename, sub=True)
# get normalized source and destination paths
dprintf(2, "data product: %s, rename %s, policy %s\n", dp.sourcepath, dp.filename, dp.policy)
sourcepath = Purr.canonizePath(dp.sourcepath)
destname = dp.fullpath = os.path.join(pathname, dp.filename)
dprintf(2, "data product: %s -> %s\n", sourcepath, destname)
# does the destination product already exist? skip if same file, else remove
if os.path.exists(destname):
if os.path.samefile(destname, sourcepath):
dprintf(2, "same file, skipping\n")
dp.timestamp = os.path.getmtime(destname)
dps.append(dp)
continue
if os.system("/bin/rm -fr '%s'" % destname):
print("Error removing %s, which is in the way of %s" % (destname, sourcepath))
print("This data product is not saved.")
continue
# for directories, compress with tar
if os.path.isdir(sourcepath):
sourcepath = sourcepath.rstrip('/')
if dp.policy == "copy" or dp.policy.startswith("move"):
dprintf(2, "archiving to tgz\n")
if os.system("tar zcf '%s' -C '%s' '%s'" % (destname,
os.path.dirname(sourcepath),
os.path.basename(sourcepath))):
print("Error archiving %s to %s" % (sourcepath, destname))
print("This data product is not saved.")
continue
if dp.policy.startswith("move"):
os.system("/bin/rm -fr '%s'" % sourcepath)
# else just a file
else:
# now copy/move it over
if dp.policy == "copy":
dprintf(2, "copying\n")
if _copy_update(sourcepath, destname):
print("Error copying %s to %s" % (sourcepath, destname))
print("This data product is not saved.")
continue
elif dp.policy.startswith('move'):
if _move_update(sourcepath, destname):
print("Error moving %s to %s" % (sourcepath, destname))
print("This data product is not saved.")
continue
# success, set timestamp and append
dp.timestamp = os.path.getmtime(destname)
dp.archived = True
dps.append(dp)
# reset list of data products
self.dps = dps
# now write out content
self.cached_include = os.path.join(pathname, 'index.include.html')
self.cached_include_valid = False
self.index_file = os.path.join(pathname, "index.html")
self.generateIndex(refresh=refresh, refresh_index=refresh_index and time.time())
self.updated = False
def addWatchedDirectory(self,
dirname,
watching=Purr.WATCHED,
save_config=True):
"""Starts watching the specified directories for changes"""
# see if we're alredy watching this exact set of directories -- do nothing if so
dirname = Purr.canonizePath(dirname)
# do nothing if already watching
if dirname in self.watched_dirs:
dprint(1, "addWatchDirectory(): already watching %s\n", dirname)
# watching=None means do not change the watch-state
if watching is None:
return
else:
if watching is None:
watching = Purr.WATCHED
# make watcher object
wdir = Purrer.WatchedDir(dirname,
mtime=self.timestamp,
watch_patterns=self._watch_patterns,
ignore_patterns=self._ignore_patterns)
# fileset=None indicates error reading directory, so ignore it
if wdir.fileset is None:
print(
"There was an error reading the directory %s, will stop watching it."
% dirname)
self.setWatchingState(dirname, Purr.REMOVED, save_config=True)
return
self.watchers[dirname] = wdir
self.watched_dirs.append(dirname)
dprintf(2, "watching directory %s, mtime %s, %d files\n", dirname,
time.strftime("%x %X", time.localtime(wdir.mtime)),
len(wdir.fileset))
# find files in this directory matching the watch_patterns, and watch them for changes
watchset = set()
for patt in self._watch_patterns:
watchset.update(fnmatch.filter(wdir.fileset, patt))
for fname in watchset:
quiet = matches_patterns(fname, self._quiet_patterns)
fullname = Purr.canonizePath(os.path.join(dirname, fname))
if fullname not in self.watchers:
wfile = Purrer.WatchedFile(fullname,
quiet=quiet,
mtime=self.timestamp)
self.watchers[fullname] = wfile
dprintf(
3, "watching file %s, timestamp %s, quiet %d\n",
fullname,
time.strftime("%x %X",
time.localtime(wfile.mtime)), quiet)
# find subdirectories matching the subdir_patterns, and watch them for changes
for fname in wdir.fileset:
fullname = Purr.canonizePath(os.path.join(dirname, fname))
if os.path.isdir(fullname):
for desc, dir_patts, canary_patts in self._subdir_patterns:
if matches_patterns(fname, dir_patts):
quiet = matches_patterns(fname,
self._quiet_patterns)
wdir = Purrer.WatchedSubdir(
fullname,
canary_patterns=canary_patts,
quiet=quiet,
mtime=self.timestamp)
self.watchers[fullname] = wdir
dprintf(
3,
"watching subdirectory %s/{%s}, timestamp %s, quiet %d\n",
fullname, ",".join(canary_patts),
time.strftime("%x %X",
time.localtime(wdir.mtime)),
quiet)
break
# set state and save config
self.setWatchingState(dirname, watching, save_config=save_config)
def _attach(self, purrlog, watchdirs=None):
"""Attaches Purr to a purrlog directory, and loads content.
Returns False if nothing new has been loaded (because directory is the same),
or True otherwise."""
purrlog = os.path.abspath(purrlog)
dprint(1, "attaching to purrlog", purrlog)
self.logdir = purrlog
self.indexfile = os.path.join(self.logdir, "index.html")
self.logtitle = "Unnamed log"
self.timestamp = self.last_scan_timestamp = time.time()
self._initIndexDir()
# reset internal state
self.ignorelistfile = None
self.autopounce = False
self.watched_dirs = []
self.entries = []
self._default_dp_props = {}
self.watchers = {}
self.temp_watchers = {}
self.attached = False
self._watching_state = {}
# check that we hold a lock on the directory
self.lockfile = os.path.join(self.logdir, ".purrlock")
# try to open lock file for r/w
try:
self.lockfile_fd = os.open(self.lockfile, os.O_RDWR | os.O_CREAT)
except:
raise Purrer.LockFailError(
"failed to open lock file %s for writing" % self.lockfile)
# try to acquire lock on the lock file
try:
fcntl.lockf(self.lockfile_fd, fcntl.LOCK_EX | fcntl.LOCK_NB)
except:
other_lock = os.fdopen(self.lockfile_fd, 'r').read()
self.lockfile_fd = None
raise Purrer.LockedError(other_lock)
# got lock, write our ID to the lock file
global _lockstring
try:
self.lockfile_fobj = os.fdopen(self.lockfile_fd, 'w')
self.lockfile_fobj.write(_lockstring)
self.lockfile_fobj.flush()
os.fsync(self.lockfile_fd)
except:
raise
# raise Purrer.LockFailError("cannot write to lock file %s"%self.lockfile)
# load log state if log directory already exists
if os.path.exists(self.logdir):
_busy = Purr.BusyIndicator()
if os.path.exists(self.indexfile):
try:
parser = Purr.Parsers.LogIndexParser()
for line in open(self.indexfile):
parser.feed(line)
self.logtitle = parser.title or self.logtitle
self.timestamp = parser.timestamp or self.timestamp
dprintf(
2, "attached log '%s', timestamp %s\n", self.logtitle,
time.strftime("%x %X", time.localtime(self.timestamp)))
except:
traceback.print_exc()
print("Error parsing %s, reverting to defaults" %
self.indexfile)
# load log entries
entries = []
for fname in os.listdir(self.logdir):
pathname = os.path.join(self.logdir, fname)
if Purr.LogEntry.isValidPathname(pathname):
try:
entry = Purr.LogEntry(load=pathname)
dprint(2, "loaded log entry", pathname)
except:
print("Error loading entry %s, skipping" % fname)
traceback.print_exc()
continue
entries.append(entry)
else:
dprint(2, fname, "is not a valid Purr entry")
# sort log entires by timestamp
entries.sort(lambda a, b: cmp(a.timestamp, b.timestamp))
self.setLogEntries(entries, save=False)
# update own timestamp
if entries:
self.timestamp = max(self.timestamp, entries[-1].timestamp)
# else logfile doesn't exist, create it
else:
self._initIndexDir()
# load configuration if it exists
# init config file
self.dirconfig = configparser.RawConfigParser()
self.dirconfigfile = os.path.join(self.logdir, "dirconfig")
if os.path.exists(self.dirconfigfile):
try:
self.dirconfig.read(self.dirconfigfile)
except:
print("Error loading config file %s" % self.dirconfigfile)
traceback.print_exc()
# load directory configuration
for dirname in self.dirconfig.sections():
try:
watching = self.dirconfig.getint(dirname, "watching")
except:
watching = Purr.WATCHED
dirname = os.path.expanduser(dirname)
self.addWatchedDirectory(dirname, watching, save_config=False)
# start watching the specified directories
for name in (watchdirs or []):
self.addWatchedDirectory(name, watching=None)
# Finally, go through list of ignored files and mark their watchers accordingly.
# The ignorelist is a list of lines of the form "timestamp filename", giving the timestamp when a
# file was last "ignored" by the purrlog user.
self.ignorelistfile = os.path.join(self.logdir, "ignorelist")
if os.path.exists(self.ignorelistfile):
# read lines from file, ignore exceptions
ignores = {}
try:
for line in open(self.ignorelistfile).readlines():
timestamp, policy, filename = line.strip().split(" ", 2)
# update dictiornary with latest timestamp
ignores[filename] = int(timestamp), policy
except:
print("Error reading %s" % self.ignorelistfile)
traceback.print_exc()
# now scan all listed files, and make sure their watchers' mtime is no older than the given
# last-ignore-timestamp. This ensures that we don't pounce on these files after restarting purr.
for filename, (timestamp, policy) in ignores.items():
watcher = self.watchers.get(filename, None)
if watcher:
watcher.mtime = max(watcher.mtime, timestamp)
# init complete
self.attached = True
return True
def regenerate(self):
Purr.progressMessage("reading %s" % self.dp.filename, sub=True)
# init fitsfile to None, so that _read() above is forced to re-read it
fitsfile = fits.open(self.dp.fullpath)
header = fitsfile[0].header
dprintf(3, "beginning render of", self.dp.fullpath);
t0 = time.time()
# write out FITS header
self.headerfile, path, uptodate = self.subproduct("-fitsheader.html")
if not uptodate:
title = "FITS header for %s" % self.dp.filename
html = """<HTML><BODY><TITLE>%s</TITLE>
<H2>%s</H2>
<PRE>""" % (title, title)
for line in header.ascard:
line = str(line).replace("<", "<").replace(">", ">")
html += line + "\n"
html += """
</PRE></BODY></HTML>\n"""
try:
open(path, "w").write(html)
except:
print("Error writing file %s" % path)
traceback.print_exc()
self.headerfile = None
# figure out number of images to include
ndim = header['NAXIS']
fitsshape = [header['NAXIS%d' % i] for i in range(1, ndim + 1)]
self.cubesize = 'x'.join(map(str, fitsshape))
if ndim < 2:
raise TypeError("can't render one-dimensional FITS files""")
elif ndim == 2:
fitsdata_to_images = lambda fdata: [fdata]
nplanes = 1
else:
ax1 = ax2 = None
# find the X/Y axes, by looking at CTYPEx
# note that the array axes are in reverse order. I.e. if X is FITS axis 1 and Y is axis 2,
# the array will be of e.g. shape 1,1,NY,NX, while fitsshape is [NX,NY,1,1]
for i in range(1, ndim + 1):
ctype = header['CTYPE%d' % i]
if [prefix for prefix in ("RA", "GLON", "ELON", "HLON", "SLON") if ctype.startswith(prefix)] \
or ctype in ("L", "X"):
ax1 = ndim - i
elif [prefix for prefix in ("DEC", "GLAT", "ELAT", "HLAT", "SLAT") if ctype.startswith(prefix)] \
or ctype in ("M", "Y"):
ax2 = ndim - i
if ax1 is None or ax2 is None:
ax1, ax2 = 1, 0
arrshape = fitsshape[-1::-1]
# this is how many planes we render, at most
nplanes = max(self.getOption('fits-nimage'), 1)
slices = []
baseslice = [0] * ndim
baseslice[ax1] = baseslice[ax2] = None
imgshape = (arrshape[min(ax1, ax2)], arrshape[max(ax1, ax2)])
while len(slices) < nplanes:
slices.append(tuple(baseslice))
for idim in range(ndim):
if baseslice[idim] != None:
baseslice[idim] += 1
if baseslice[idim] < arrshape[idim]:
break
else:
baseslice[idim] = 0
else:
break
nplanes = len(slices)
# OK, slices contains how many slices to return
def fitsdata_to_images(fdata, slices=slices, imgshape=imgshape):
dprint(3, "fitsdata_to_images", slices, fdata.shape);
t0 = time.time()
# reshape to collapse into a 3D cube
img = [fdata[i].reshape(imgshape) for i in slices]
dprint(3, "collecting images took", time.time() - t0, "secs");
t0 = time.time()
return img
# OK, now cycle over all images
dprintf(3, "%s: rendering %d planes\n", self.dp.fullpath, nplanes);
t0 = time.time()
self.imgrec = [None] * nplanes
# get number of bins (0 or None means no histogram)
nbins = self.getOption("fits-hist-nbin")
# see if histogram clipping is enabled, set hclip to None if not
self.hclip = hclip = self.getOption("fits-hist-clip")
if hclip == 1 or not nbins:
hclip = None
tsize_img = self.getOption("image-thumbnail-width"), self.getOption("image-thumbnail-height")
tsize_hist = self.getOption("hist-thumbnail-width"), self.getOption("hist-thumbnail-height")
self.hist_size = self.getOption("hist-width"), self.getOption("hist-height")
# filled once we read the data
images = None
for num_image in range(nplanes):
# do we have a cached status record for this image?
recfile, recpath, uptodate = self.subproduct("-%d-stats" % num_image)
if uptodate:
dprintf(3, "%s(%d): stats file %s up-to-date, reading in\n", self.dp.fullpath, num_image, recfile)
try:
self.imgrec[num_image] = pickle.load(file(recpath))
continue
except:
print("Error reading stats file %s, regenerating everything" % recpath)
traceback.print_exc()
# out of date, so we regenerate everything
# build up record of stuff associated with this image
rec = self.imgrec[num_image] = Kittens.utils.recdict()
# generate paths for images
rec.fullimage, img_path = self.subproductPath("-%d-full.png" % num_image)
rec.thumbnail, img_thumb = self.subproductPath("-%d-thumb.png" % num_image)
if pychart:
rec.histogram_full, hf_path = self.subproductPath("-%d-hist-full.png" % num_image)
rec.histogram_zoom, hz_path = self.subproductPath("-%d-hist-zoom.png" % num_image)
rec.histogram_full_thumb, hf_thumb = self.subproductPath("-%d-hist-full-thumb.png" % num_image)
rec.histogram_zoom_thumb, hz_thumb = self.subproductPath("-%d-hist-zoom-thumb.png" % num_image)
# need to read in data at last
if not images:
dprint(3, "reading data");
t0 = time.time()
fitsdata = fitsfile[0].data
dprint(3, "reading data took", time.time() - t0, "secs");
t0 = time.time()
fitsfile = None
images = fitsdata_to_images(fitsdata)
dprint(3, "converting to images took", time.time() - t0, "secs");
t0 = time.time()
fitsdata = None
data = images[num_image]
title = self.dp.filename
if nplanes > 1:
title += ", plane #%d" % num_image
Purr.progressMessage("rendering %s" % title, sub=True)
# min/max data values
dprint(3, "rendering plane", num_image);
t0 = time.time()
datamask = ~numpy.isfinite(data)
dprint(3, "making mask took", time.time() - t0, "secs");
t0 = time.time()
datamin, datamax = scipy.ndimage.measurements.extrema(data, datamask, False)[:2]
dprint(3, "computing min/max took", time.time() - t0, "secs");
t0 = time.time()
rec.datamin, rec.datamax = datamin, datamax
# mean and sigma
rec.datamean = scipy.ndimage.measurements.mean(data, datamask, False)
dprint(3, "computing mean took", time.time() - t0, "secs");
t0 = time.time()
rec.datastd = scipy.ndimage.measurements.standard_deviation(data, datamask, False)
dprint(3, "computing std took", time.time() - t0, "secs");
t0 = time.time()
# thumbnail files will be "" if images are small enough to be inlined.
# these will be None if no histogram clipping is applied
rec.clipmin, rec.clipmax = None, None
dprintf(3, "%s plane %d: datamin %g, datamax %g\n", self.dp.fullpath, num_image, rec.datamin, rec.datamax)
# compute histogram of data only if this is enabled,
# and either pychart is available (so we can produce plots), or histogram clipping is in effect
if datamin != datamax and nbins and (pychart or hclip):
dprintf(3, "%s plane %d: computing histogram\n", self.dp.fullpath, num_image)
counts = scipy.ndimage.measurements.histogram(data, datamin, datamax, nbins, labels=datamask,
index=False); # needed for 1.3+ to avoid warnings
edges = datamin + (datamax - datamin) * (numpy.arange(nbins, dtype=float) + .5) / nbins
dprint(3, "computing histogram took", time.time() - t0, "secs");
t0 = time.time()
# render histogram
if pychart:
try:
self._make_histogram(hf_path, "Histogram of %s" % title, edges, counts)
dprint(3, "rendering histogram took", time.time() - t0, "secs");
t0 = time.time()
except:
print("Error rendering histogram %s" % hf_path)
traceback.print_exc()
rec.histogram_full = None
# if histogram was rendered, make a thumbnail
if rec.histogram_full:
self.makeThumb(hf_path, hf_thumb, tsize_hist)
else:
rec.histogram_full_thumb = None
# now, compute clipped data if needed
if hclip:
# find max point in histogram
ic = counts.argmax()
# compute number of points that need to be included, given the clip factor
target_count = int(data.size * hclip)
ih0 = ih1 = ic
totcount = counts[ic]
# find how many bins to include around ic, stopping when we hit the edge
while totcount < target_count:
if ih0 > 0:
ih0 -= 1
totcount += counts[ih0]
if ih1 < nbins - 1:
ih1 += 1
totcount += counts[ih1]
# just in case
if ih0 <= 0 and ih1 >= nbins - 1:
break
# and these are the clipping limits
datamin = float(edges[ih0])
if ih1 >= nbins - 1:
ih1 = nbins - 1; # and datamax is already the clipping limit
else:
ih1 += 1
datamax = float(edges[ih1])
rec.clipmin, rec.clipmax = datamin, datamax
dprintf(3, "%s plane %d: clipping to %g,%g\n", self.dp.fullpath, num_image, rec.clipmin,
rec.clipmax)
# render zoomed histogram
if pychart:
if rec.clipmax != rec.clipmin:
zcounts = scipy.ndimage.measurements.histogram(data, rec.clipmin, rec.clipmax, nbins,
labels=datamask,
index=False); # needed for 1.3+ to avoid warnings
zedges = rec.clipmin + (rec.clipmax - rec.clipmin) * (
numpy.arange(nbins, dtype=float) + .5) / nbins
try:
self._make_histogram(hz_path, "Histogram zoom of %s" % title, zedges, zcounts)
dprint(3, "rendering zoomed histogram took", time.time() - t0, "secs");
t0 = time.time()
except:
print("Error rendering histogram %s" % hz_path)
traceback.print_exc()
rec.histogram_zoom = None
else: # no meaningful zoomed area to render
rec.histogram_zoom = None
# if histogram was rendered, make a thumbnail
if rec.histogram_zoom:
histogram_zoom_thumb = self.makeThumb(hz_path, hz_thumb, tsize_hist)
else:
rec.histogram_zoom_thumb = None
# clip data
data = numpy.clip(data, datamin, datamax)
# end of clipping
# else no histogram for whatever reason
else:
rec.histogram_full = rec.histogram_zoom = rec.histogram_full_thumb = rec.histogram_zoom_thumb = None
# ok, data has been clipped if need be. Rescale it to 8-bit integers
t0 = time.time()
datarng = datamax - datamin
if datarng:
data = (data - datamin) * (255 / datarng)
data = data.round().astype('uint8')
data[datamask] = 255
else:
data = numpy.zeros(data.shape, dtype='uint8')
dprintf(3, "%s plane %d: rescaled to %d:%d in %f seconds\n", self.dp.fullpath, num_image, data.min(),
data.max(), time.time() - t0);
t0 = time.time()
# generate PNG image
img = None
try:
img = PIL.Image.frombuffer('L', data.shape[-1::-1], numpy.getbuffer(data), "raw", 'L', 0, -1)
dprint(3, "image frombuffer took", time.time() - t0, "secs");
t0 = time.time()
# img = PIL.Image.new('L',data.shape)
# dprint(3,"new image took",time.time()-t0,"secs"); t0 = time.time()
# imgdata = data.reshape((data.size,))
# dprint(3,"data.reshape took",time.time()-t0,"secs"); t0 = time.time()
# img.putdata(imgdata)
# dprint(3,"putdata took",time.time()-t0,"secs"); t0 = time.time()
# img = img.transpose(PIL.Image.FLIP_TOP_BOTTOM)
# dprint(3,"transpose took",time.time()-t0,"secs"); t0 = time.time()
img.save(img_path, 'PNG')
dprint(3, "saving took", time.time() - t0, "secs");
t0 = time.time()
except:
print("Error rendering image %s" % path)
traceback.print_exc()
rec.fullimage = img = None
# if image was rendered, make a thumbnail
if rec.fullimage:
thumb = self.makeThumb(img_path, img_thumb, tsize_img, img=img)
dprint(3, "rendering thumbnail took", time.time() - t0, "secs");
t0 = time.time()
# None means thumbnail failed
if thumb is None:
rec.thumbnail = None
# else perhaps image is its own thumbnail
elif thumb is img_path:
rec.thumbnail = rec.fullimage
else:
rec.thumbnail = None
# write stats
try:
pickle.dump(rec, file(recpath, 'w'))
except:
print("Error writing stats file %s" % recpath)
traceback.print_exc()
def restore_from_archive(self, parent=None):
"""Function to restore a DP from archived copy
Asks for confirmation along the way if parent is not None
(in which case it will be the parent widget for confirmation dialogs)
"""
from PyQt4.Qt import QMessageBox
exists = os.path.exists(self.sourcepath)
if parent:
msg = """<P>Do you really want to restore <tt>%s</tt> from
this entry's copy of <tt>%s</tt>?</P>""" % (self.sourcepath,
self.filename)
exists = os.path.exists(self.sourcepath)
if exists:
msg += """<P>Current file exists, and will be overwritten.</P>"""
if QMessageBox.warning(parent, "Restoring from archive", msg,
QMessageBox.Yes,
QMessageBox.No) != QMessageBox.Yes:
return False
else:
if QMessageBox.question(parent, "Restoring from archive", msg,
QMessageBox.Yes,
QMessageBox.No) != QMessageBox.Yes:
return False
busy = Purr.BusyIndicator()
# remove file if in the way
if exists:
if os.system("/bin/rm -fr '%s'" % self.sourcepath):
busy = None
if parent:
QMessageBox.warning(
parent, "Error removing file", """<P>
There was an error removing %s. Archived copy was not restored.
The text console may have more information.</P>""" %
self.sourcepath, QMessageBox.Ok, 0)
return False
# unpack archived file
if self.fullpath.endswith('.tgz'):
parent_dir = os.path.dirname(self.sourcepath.rstrip('/'))
os.system("/bin/rm -fr %s" % self.sourcepath)
if os.system("tar zxf '%s' -C '%s'" % (self.fullpath, parent_dir)):
busy = None
if parent:
QMessageBox.warning(
parent, "Error unpacking file", """<P>
There was an error unpacking the archived version to %s. The text console may have more information.</P>"""
% self.sourcepath, QMessageBox.Ok, 0)
return False
# else simply copy over
else:
if os.system("/bin/cp -a '%s' '%s'" %
(self.fullpath, self.sourcepath)):
busy = None
if parent:
QMessageBox.warning(
parent, "Error copying file", """<P>
There was an error copying the archived version to %s. The text console may have more information.</P>"""
% self.sourcepath, QMessageBox.Ok, 0)
return False
busy = None
if parent:
QMessageBox.information(
parent, "Restored file", """<P>Restored %s from this entry's
archived copy.</P>""" % self.sourcepath, QMessageBox.Ok, 0)
return True
# need to read in data at last
if not images:
dprint(3,"reading data"); t0 = time.time();
fitsdata = fitsfile[0].data;
dprint(3,"reading data took",time.time()-t0,"secs"); t0 = time.time();
fitsfile = None;
images = fitsdata_to_images(fitsdata);
dprint(3,"converting to images took",time.time()-t0,"secs"); t0 = time.time();
fitsdata = None;
data = images[num_image];
title = self.dp.filename;
if nplanes > 1:
title += ", plane #%d"%num_image;
Purr.progressMessage("rendering %s"%title,sub=True);
# min/max data values
dprint(3,"rendering plane",num_image); t0 = time.time();
datamask = ~numpy.isfinite(data);
dprint(3,"making mask took",time.time()-t0,"secs"); t0 = time.time();
datamin,datamax = scipy.ndimage.measurements.extrema(data,datamask,False)[:2];
dprint(3,"computing min/max took",time.time()-t0,"secs"); t0 = time.time();
rec.datamin,rec.datamax = datamin,datamax;
# mean and sigma
rec.datamean = scipy.ndimage.measurements.mean(data,datamask,False);
dprint(3,"computing mean took",time.time()-t0,"secs"); t0 = time.time();
rec.datastd = scipy.ndimage.measurements.standard_deviation(data,datamask,False);
dprint(3,"computing std took",time.time()-t0,"secs"); t0 = time.time();
# thumbnail files will be "" if images are small enough to be inlined.
# these will be None if no histogram clipping is applied
def save(self,
dirname=None,
refresh=0,
refresh_index=True,
emit_message=True):
"""Saves entry in the given directory. Data products will be copied over if not
residing in that directory.
'refresh' is a timestamp, passed to renderIndex(), causing all data products OLDER than the specified time to be regenerated.
'refresh_index', if true, causes index files to be re-rendered unconditionally
"""
if not refresh and not self.updated:
return
timestr = time.strftime("%Y%m%d-%H%M%S",
time.localtime(self.timestamp))
Purr.progressMessage("Rendering entry for %s" % timestr)
if dirname:
self.pathname = pathname = os.path.join(
dirname,
"%s-%s" % (("ignore" if self.ignore else "entry"), timestr))
elif not self.pathname:
raise ValueError("Cannot save entry: pathname not specified")
else:
pathname = self.pathname
# set timestamp
if not self.timestamp:
self.timestamp = int(time.time())
# get canonized path to output directory
pathname = Purr.canonizePath(pathname)
if not os.path.exists(pathname):
os.mkdir(pathname)
# now save content
# get device of pathname -- need to know whether we move or copy
devnum = os.stat(pathname).st_dev
# copy data products as needed
dprintf(2, "saving entry %s, %d data products\n", pathname,
len(self.dps))
dps = []
for dp in self.dps:
# if archived, this indicates a previously saved data product, so ignore it
# if ignored, no need to save the DP -- but keep it in list
if dp.archived or dp.ignored:
dprintf(3, "dp %s is archived or ignored, skipping\n",
dp.sourcepath)
dps.append(dp)
continue
# file missing for some reason (perhaps it got removed on us?) skip data product entirely
if not os.path.exists(dp.sourcepath):
dprintf(2, "data product %s missing, ignoring\n",
dp.sourcepath)
continue
Purr.progressMessage("archiving %s" % dp.filename, sub=True)
# get normalized source and destination paths
dprintf(2, "data product: %s, rename %s, policy %s\n",
dp.sourcepath, dp.filename, dp.policy)
sourcepath = Purr.canonizePath(dp.sourcepath)
destname = dp.fullpath = os.path.join(pathname, dp.filename)
dprintf(2, "data product: %s -> %s\n", sourcepath, destname)
# does the destination product already exist? skip if same file, else remove
if os.path.exists(destname):
if os.path.samefile(destname, sourcepath):
dprintf(2, "same file, skipping\n")
dp.timestamp = os.path.getmtime(destname)
dps.append(dp)
continue
if os.system("/bin/rm -fr '%s'" % destname):
print("Error removing %s, which is in the way of %s" %
(destname, sourcepath))
print("This data product is not saved.")
continue
# for directories, compress with tar
if os.path.isdir(sourcepath):
sourcepath = sourcepath.rstrip('/')
if dp.policy == "copy" or dp.policy.startswith("move"):
dprintf(2, "archiving to tgz\n")
if os.system("tar zcf '%s' -C '%s' '%s'" %
(destname, os.path.dirname(sourcepath),
os.path.basename(sourcepath))):
print("Error archiving %s to %s" %
(sourcepath, destname))
print("This data product is not saved.")
continue
if dp.policy.startswith("move"):
os.system("/bin/rm -fr '%s'" % sourcepath)
# else just a file
else:
# now copy/move it over
if dp.policy == "copy":
dprintf(2, "copying\n")
if _copy_update(sourcepath, destname):
print("Error copying %s to %s" %
(sourcepath, destname))
print("This data product is not saved.")
continue
elif dp.policy.startswith('move'):
if _move_update(sourcepath, destname):
print("Error moving %s to %s" % (sourcepath, destname))
print("This data product is not saved.")
continue
# success, set timestamp and append
dp.timestamp = os.path.getmtime(destname)
dp.archived = True
dps.append(dp)
# reset list of data products
self.dps = dps
# now write out content
self.cached_include = os.path.join(pathname, 'index.include.html')
self.cached_include_valid = False
self.index_file = os.path.join(pathname, "index.html")
self.generateIndex(refresh=refresh,
refresh_index=refresh_index and time.time())
self.updated = False
