# http://pyrocko.org - GPLv3
#
# The Pyrocko Developers, 21st Century
# ---|P------/S----------~Lg----------
'''
Squirrel core file reading and indexing.
'''
import time
import logging
from pyrocko import util
from pyrocko.io.io_common import FileLoadError
from pyrocko import progress
from .backends import \
mseed, sac, datacube, stationxml, textfiles, virtual, yaml, tdms_idas, \
spickle
from ..model import to_kind_ids, EMPTY, Nut
backend_modules = [
mseed, sac, datacube, stationxml, textfiles, virtual, yaml, tdms_idas,
spickle]
logger = logging.getLogger('psq.io')
def make_task(*args):
return progress.task(*args, logger=logger)
g_format_providers = {}
update_format_providers()
[docs]def get_backend(fmt):
'''
Get squirrel io backend module for a given file format.
:param fmt:
Format identifier.
:type fmt:
str
'''
try:
return g_format_providers[fmt][0]
except KeyError:
raise UnknownFormat(fmt)
g_content_kinds = ['waveform', 'station', 'channel', 'response', 'event']
[docs]def supported_content_kinds():
'''
Get list of supported content kinds offered through Squirrel.
'''
return g_content_kinds + ['waveform_promise']
[docs]def iload(
paths,
segment=None,
format='detect',
database=None,
check=True,
skip_unchanged=False,
content=g_content_kinds,
show_progress=True,
update_selection=None):
'''
Iteratively load content or index/reindex meta-information from files.
:param paths:
Iterator yielding file names to load from or a Squirrel selection
object providing the file names.
:type paths:
iterator yielding :py:class:`str` or
:py:class:`~pyrocko.squirrel.selection.Selection`
:param segment:
File-specific segment identifier (can only be used when loading from a
single file).
:type segment:
int
:param format:
File format identifier or ``'detect'`` for autodetection. When loading
from a selection, per-file format assignation is taken from the hint in
the selection and this flag is ignored.
:type format:
str
:param database:
Database to use for meta-information caching. When loading from a
selection, this should be ``None`` and the database from the selection
is used.
:type database:
:py:class:`~pyrocko.squirrel.database.Database`
:param check:
If ``True``, investigate modification time and file sizes of known
files to debunk modified files (pessimistic mode), or ``False`` to
deactivate checks (optimistic mode).
:type check:
bool
:param skip_unchanged:
If ``True``, only yield index nuts for new / modified files.
:type skip_unchanged:
bool
:param content:
Selection of content types to load.
:type content:
:py:class:`list` of :py:class:`str`
This generator yields :py:class:`~pyrocko.squirrel.model.Nut` objects for
individual pieces of information found when reading the given files. Such a
nut may represent a waveform, a station, a channel, an event or other data
type. The nut itself only contains the meta-information. The actual content
information is attached to the nut if requested. All nut meta-information
is stored in the squirrel meta-information database. If possible, this
function avoids accessing the actual disk files and provides the requested
information straight from the database. Modified files are recognized and
reindexed as needed.
'''
from ..selection import Selection
n_db = 0
n_load = 0
selection = None
kind_ids = to_kind_ids(content)
if isinstance(paths, str):
paths = [paths]
else:
if segment is not None:
raise TypeError(
'iload: segment argument can only be used when loading from '
'a single file')
if isinstance(paths, Selection):
selection = paths
if database is not None:
raise TypeError(
'iload: database argument must be None when called with a '
'selection')
database = selection.get_database()
if skip_unchanged and not isinstance(paths, Selection):
raise TypeError(
'iload: need selection when called with "skip_unchanged=True"')
temp_selection = None
transaction = None
if database:
if not selection:
# Avoid creating temporary selection for small batches.
# this is helpful because then, we can avoid locking the database,
# e.g. during loading of content, when the content has not been
# modified.
paths = util.short_to_list(100, paths)
if isinstance(paths, list) and len(paths) == 0:
return
if not (isinstance(paths, list) and len(paths) < 100
and not skip_unchanged):
temp_selection = database.new_selection(
paths, show_progress=show_progress, format=format)
selection = temp_selection
if skip_unchanged:
selection.flag_modified(check)
if selection:
# undig_grouped starts a long select which causes deadlocks
# when transaction is started after starting the select, therefore
# the transaction has to be started before in these cases.
# The db will be locked for a long time in this case. This could be
# solved either by breaking the indexing into smaller blocks in
# the caller or by modifying undig_grouped to allow limit and
# offset and add an outer loop below.
transaction = database.transaction(
'update content index')
transaction.begin()
it = selection.undig_grouped(skip_unchanged=skip_unchanged)
else:
# The list() causes the query to finish, so we don't have to lock,
# and can start a transaction only when encountering a modified/new
# file.
it = list(database.undig_few(paths, format=format))
else:
it = (((format, path), []) for path in paths)
it = util.short_to_list(100, iter(it))
try:
n_files_total = len(it)
if n_files_total == 0:
if transaction:
transaction.commit()
transaction.close()
return
except TypeError:
n_files_total = None
task = None
if show_progress:
if not kind_ids:
task = make_task('Indexing files', n_files_total)
else:
task = make_task('Loading files', n_files_total)
n_files = 0
tcommit = time.time()
clean = False
try:
for (format, path), old_nuts in it:
if task is not None:
condition = '(nuts: %i from file, %i from cache)\n %s' % (
n_load, n_db, path)
task.update(n_files, condition)
n_files += 1
# cannot release when iterating a selection (see above)
if database and transaction and not selection:
tnow = time.time()
if tnow - tcommit > 20. or n_files % 1000 == 0:
transaction.commit()
tcommit = tnow
transaction.begin()
try:
if check and old_nuts and old_nuts[0].file_modified():
old_nuts = []
modified = True
else:
modified = False
if segment is not None:
old_nuts = [
nut for nut in old_nuts if nut.file_segment == segment]
if old_nuts:
db_only_operation = not kind_ids or all(
nut.kind_id in kind_ids and nut.content_in_db
for nut in old_nuts)
if db_only_operation:
# logger.debug('using cached information for file %s, '
# % path)
for nut in old_nuts:
if nut.kind_id in kind_ids:
database.undig_content(nut)
n_db += 1
yield nut
continue
if format == 'detect':
if old_nuts and not old_nuts[0].file_modified():
format_this = old_nuts[0].file_format
else:
format_this = detect_format(path)
else:
format_this = format
mod = get_backend(format_this)
mtime, size = mod.get_stats(path)
if segment is not None:
logger.debug(
'Reading file "%s", segment "%s".' % (path, segment))
else:
logger.debug(
'Reading file "%s".' % path)
nuts = []
for nut in mod.iload(format_this, path, segment, content):
nut.file_path = path
nut.file_format = format_this
nut.file_mtime = mtime
nut.file_size = size
if nut.content is not None:
nut.content._squirrel_key = nut.key
nuts.append(nut)
n_load += 1
yield nut
if segment is None and len(nuts) == 0:
nuts.append(
Nut(
file_path=path,
file_format=format_this,
file_mtime=mtime,
file_size=size,
kind_id=EMPTY))
if database and nuts != old_nuts:
if old_nuts or modified:
logger.debug(
'File has been modified since last access: %s'
% path)
if segment is not None:
nuts = list(mod.iload(format_this, path, None, []))
for nut in nuts:
nut.file_path = path
nut.file_format = format_this
nut.file_mtime = mtime
nut.file_size = size
if len(nuts) == 0:
nuts.append(
Nut(
file_path=path,
file_format=format_this,
file_mtime=mtime,
file_size=size,
kind_id=EMPTY))
if not transaction:
transaction = database.transaction(
'update content index')
transaction.begin()
database.dig(nuts, transaction=transaction)
if update_selection is not None:
update_selection._set_file_states_force_check(
[path], transaction=transaction)
update_selection._update_nuts(transaction=transaction)
except FileLoadError:
logger.error('Cannot read file: %s' % path)
if database:
if not transaction:
transaction = database.transaction(
'update content index')
transaction.begin()
database.reset(path, transaction=transaction)
clean = True
finally:
if task is not None:
condition = '(nuts: %i from file, %i from cache)' % (n_load, n_db)
task.update(n_files, condition)
if clean:
task.done(condition)
else:
task.fail(condition + ' terminated')
if database and transaction:
transaction.commit()
transaction.close()
if temp_selection:
del temp_selection
__all__ = [
'iload',
'detect_format',
'supported_formats',
'supported_content_kinds',
'get_backend',
'FormatDetectionFailed',
'UnknownFormat',
]
