def view_results(request, id, first_row=0):
"""
Returns the view for the results of the QueryHistory with the given id.
The query results MUST be ready.
To display query results, one should always go through the watch_query view.
If ``first_row`` is 0, restarts (if necessary) the query read. Otherwise, just
spits out a warning if first_row doesn't match the servers conception.
Multiple readers will produce a confusing interaction here, and that's known.
It understands the ``context`` GET parameter. (See watch_query().)
"""
# Coerce types; manage arguments
id = int(id)
first_row = long(first_row)
start_over = (first_row == 0)
# Retrieve models from database
query_history = models.QueryHistory.objects.get(id=id)
handle = QueryHandle(id=query_history.server_id, log_context=query_history.log_context)
context = _parse_query_context(request.GET.get('context'))
# Retrieve query results
try:
results = db_utils.db_client().fetch(handle, start_over)
assert results.ready, 'Trying to display result that is not yet ready. Query id %s' % (id,)
# We display the "Download" button only when we know
# that there are results:
downloadable = (first_row > 0 or len(results.data) > 0)
fetch_error = False
except BeeswaxException, ex:
fetch_error = True
error_message, log = expand_exception(ex)
def explain_directly(request, query_str, query_msg, design):
"""
Runs explain query.
"""
explanation = db_utils.db_client().explain(query_msg)
context = ("design", design)
return render('explain.mako', request,
dict(query=query_str, explanation=explanation.textual, query_context=context))
def watch_query(request, id):
"""
Wait for the query to finish and (by default) displays the results of query id.
It understands the optional GET params:
on_success_url
If given, it will be displayed when the query is successfully finished.
Otherwise, it will display the view query results page by default.
context
A string of "name:data" that describes the context
that generated this query result. It may be:
- "table":"<table_name>"
- "design":<design_id>
All other GET params will be passed to on_success_url (if present).
"""
# Coerce types; manage arguments
id = int(id)
query_history = authorized_get_history(request, id, must_exist=True)
# GET param: context.
context_param = request.GET.get('context', '')
# GET param: on_success_url. Default to view_results
results_url = urlresolvers.reverse(view_results, kwargs=dict(id=str(id), first_row=0, last_result_len=0))
on_success_url = request.GET.get('on_success_url')
if not on_success_url:
on_success_url = results_url
# Get the server_id
server_id, state = _get_server_id_and_state(query_history)
query_history.save_state(state)
# Query finished?
if state == QueryHistory.STATE.expired:
raise PopupException(_("The result of this query has expired."))
elif state == QueryHistory.STATE.available:
return format_preserving_redirect(request, on_success_url, request.GET)
elif state == QueryHistory.STATE.failed:
# When we fetch, Beeswax server will throw us a BeeswaxException, which has the
# log we want to display.
return format_preserving_redirect(request, results_url, request.GET)
# Still running
log = db_utils.db_client(query_history.get_query_server()).get_log(server_id)
# Keep waiting
# - Translate context into something more meaningful (type, data)
context = _parse_query_context(context_param)
return render('watch_wait.mako', request, {
'query': query_history,
'fwd_params': request.GET.urlencode(),
'log': log,
'hadoop_jobs': _parse_out_hadoop_jobs(log),
'query_context': context,
})
def data_generator(query_model, formatter):
"""
data_generator(query_model, formatter) -> generator object
Return a generator object for a csv. The first line is the column names.
This is similar to export_csvxls.generator, but has
one or two extra complexities.
"""
global _DATA_WAIT_SLEEP
is_first_row = True
next_row = 0
results = None
handle = QueryHandle(query_model.server_id, query_model.log_context)
yield formatter.init_doc()
while True:
# Make sure that we have the next batch of ready results
while results is None or not results.ready:
results = db_utils.db_client().fetch(handle,
start_over=is_first_row)
if not results.ready:
time.sleep(_DATA_WAIT_SLEEP)
# Someone is reading the results concurrently. Abort.
# But unfortunately, this current generator will produce incomplete data.
if next_row != results.start_row:
msg = 'Error: Potentially incomplete results as an error occur during data retrieval.'
yield formatter.format_row([msg])
err = ('Detected another client retrieving results for %s. '
'Expect next row being %s and got %s. Aborting' %
(query_model.server_id, next_row, results.start_row))
LOG.error(err)
raise RuntimeError(err)
if is_first_row:
is_first_row = False
yield formatter.format_header(results.columns)
else:
for i, row in enumerate(results.data):
# TODO(bc): Hive seems to always return tab delimited row data.
# What if a cell has a tab?
row = row.split('\t')
try:
yield formatter.format_row(row)
except TooBigToDownloadException, ex:
LOG.error(ex)
# Exceeded limit. Stop.
results.has_more = False
break
if results.has_more:
next_row += len(results.data)
results = None
else:
yield formatter.fini_doc()
break
def data_generator(query_model, formatter):
"""
data_generator(query_model, formatter) -> generator object
Return a generator object for a csv. The first line is the column names.
This is similar to export_csvxls.generator, but has
one or two extra complexities.
"""
global _DATA_WAIT_SLEEP
is_first_row = True
next_row = 0
results = None
handle = QueryHandle(query_model.server_id, query_model.log_context)
yield formatter.init_doc()
while True:
# Make sure that we have the next batch of ready results
while results is None or not results.ready:
results = db_utils.db_client(query_model.get_query_server()).fetch(handle, start_over=is_first_row, fetch_size=-1)
if not results.ready:
time.sleep(_DATA_WAIT_SLEEP)
# Someone is reading the results concurrently. Abort.
# But unfortunately, this current generator will produce incomplete data.
if next_row != results.start_row:
msg = _('Error: Potentially incomplete results as an error occurred during data retrieval.')
yield formatter.format_row([msg])
err = (_('Detected another client retrieving results for %(server_id)s. '
'Expected next row to be %(row)s and got %(start_row)s. Aborting') %
{'server_id': query_model.server_id, 'row': next_row, 'start_row': results.start_row})
LOG.error(err)
raise RuntimeError(err)
if is_first_row:
is_first_row = False
yield formatter.format_header(results.columns)
else:
for i, row in enumerate(results.data):
# TODO(bc): Hive seems to always return tab delimited row data.
# What if a cell has a tab?
row = row.split('\t')
try:
yield formatter.format_row(row)
except TooBigToDownloadException, ex:
LOG.error(ex)
# Exceeded limit. Stop.
results.has_more = False
break
if results.has_more:
next_row += len(results.data)
results = None
else:
yield formatter.fini_doc()
break
def expand_exception(exc):
"""expand_exception(exc) -> (error msg, log message)"""
try:
log = db_utils.db_client().get_log(exc.log_context)
except:
# Always show something, even if server has died on the job.
log = "Could not retrieve log."
if not exc.message:
error_message = "Unknown exception."
else:
error_message = exc.message
return error_message, log
def expand_exception(exc):
"""expand_exception(exc) -> (error msg, log message)"""
try:
log = db_utils.db_client().get_log(exc.log_context)
except:
# Always show something, even if server has died on the job.
log = _("Could not retrieve log.")
if not exc.message:
error_message = _("Unknown exception.")
else:
error_message = force_unicode(exc.message, strings_only=True, errors='replace')
return error_message, log
def configuration(request):
if request.method == 'POST':
server_form = QueryServerForm(request.POST)
if server_form.is_valid():
query_server = db_utils.get_query_server(server_form.cleaned_data["server"])
config_values = db_utils.db_client(query_server).get_default_configuration(
bool(request.REQUEST.get("include_hadoop", False)))
else:
server_form = QueryServerForm()
config_values = {}
return render("configuration.mako", request, {'config_values': config_values,
'server_form': server_form})
def view_results(request, id, first_row=0, last_result_len=0):
"""
Returns the view for the results of the QueryHistory with the given id.
The query results MUST be ready.
To display query results, one should always go through the watch_query view.
If ``first_row`` is 0, restarts (if necessary) the query read. Otherwise, just
spits out a warning if first_row doesn't match the servers conception.
Multiple readers will produce a confusing interaction here, and that's known.
It understands the ``context`` GET parameter. (See watch_query().)
"""
# Coerce types; manage arguments
id = int(id)
first_row = long(first_row)
start_over = (first_row == 0)
query_history = authorized_get_history(request, id, must_exist=True)
handle = QueryHandle(id=query_history.server_id,
log_context=query_history.log_context)
context = _parse_query_context(request.GET.get('context'))
# Retrieve query results
try:
results = db_utils.db_client(query_history.get_query_server()).fetch(
handle, start_over, -1)
assert results.ready, _(
'Trying to display result that is not yet ready. Query id %(id)s'
) % {
'id': id
}
# We display the "Download" button only when we know
# that there are results:
downloadable = (first_row > 0 or len(results.data) > 0)
fetch_error = False
except BeeswaxException, ex:
fetch_error = True
error_message, log = expand_exception(ex)
def save_results(request, id):
"""
Save the results of a query to an HDFS directory
"""
id = int(id)
query_history = models.QueryHistory.objects.get(id=id)
if query_history.owner != request.user:
raise PopupException('This action is only available to the user who submitted the query.')
_, state = _get_server_id_and_state(query_history)
query_history.save_state(state)
error_msg, log = None, None
if request.method == 'POST':
# Make sure the result is available.
# Note that we may still hit errors during the actual save
if state != models.QueryHistory.STATE.available:
if state in (models.QueryHistory.STATE.failed, models.QueryHistory.STATE.expired):
msg = 'This query has %s. Results unavailable.' % (state,)
else:
msg = 'The result of this query is not available yet.'
raise PopupException(msg)
form = beeswax.forms.SaveResultsForm(request.POST)
# Cancel goes back to results
if request.POST.get('cancel'):
return format_preserving_redirect(request, '/beeswax/watch/%s' % (id,))
if form.is_valid():
# Do save
# 1. Get the results metadata
assert request.POST.get('save')
handle = QueryHandle(id=query_history.server_id, log_context=query_history.log_context)
try:
result_meta = db_utils.db_client().get_results_metadata(handle)
except QueryNotFoundException, ex:
LOG.exception(ex)
raise PopupException('Cannot find query.')
if result_meta.table_dir:
result_meta.table_dir = request.fs.urlsplit(result_meta.table_dir)[2]
# 2. Check for partitioned tables
if result_meta.table_dir is None:
raise PopupException(
'Saving results from a partitioned table is not supported. '
'You may copy from the HDFS location manually.')
# 3. Actual saving of results
try:
if form.cleaned_data['save_target'] == form.SAVE_TYPE_DIR:
# To dir
if result_meta.in_tablename:
raise PopupException(
'Saving results from a table to a directory is not supported. '
'You may copy from the HDFS location manually.')
target_dir = form.cleaned_data['target_dir']
request.fs.rename_star(result_meta.table_dir, target_dir)
LOG.debug("Moved results from %s to %s" % (result_meta.table_dir, target_dir))
query_history.save_state(models.QueryHistory.STATE.expired)
fb_url = location_to_url(request, target_dir, strict=False)
popup = PopupWithJframe('Query results stored in %s' % (target_dir,),
launch_app_name='FileBrowser',
launch_app_url=fb_url)
return render_injected(list_query_history(request), popup)
elif form.cleaned_data['save_target'] == form.SAVE_TYPE_TBL:
# To new table
try:
return _save_results_ctas(request,
query_history,
form.cleaned_data['target_table'],
result_meta)
except BeeswaxException, bex:
LOG.exception(bex)
error_msg, log = expand_exception(bex)
except IOError, ex:
LOG.exception(ex)
error_msg = str(ex)
def configuration(request):
config_values = db_utils.db_client().get_default_configuration(
bool(request.REQUEST.get("include_hadoop", False)))
return render("configuration.mako", request, dict(config_values=config_values))
fetch_error = True
error_message, log = expand_exception(ex)
# Handle errors
if fetch_error:
return render('watch_results.mako', request, {
'query': query_history,
'error': True,
'error_message': error_message,
'log': log,
'hadoop_jobs': _parse_out_hadoop_jobs(log),
'query_context': context,
'can_save': False,
})
log = db_utils.db_client().get_log(query_history.server_id)
download_urls = {}
if downloadable:
for format in common.DL_FORMATS:
download_urls[format] = urlresolvers.reverse(
download, kwargs=dict(id=str(id), format=format))
save_form = beeswax.forms.SaveResultsForm()
# Display the results
return render('watch_results.mako', request, {
'error': False,
'query': query_history,
# Materialize, for easier testability.
'results': list(parse_results(results.data)),
'has_more': results.has_more,
def watch_query(request, id):
"""
Wait for the query to finish and (by default) displays the results of query id.
It understands the optional GET params:
on_success_url
If given, it will be displayed when the query is successfully finished.
Otherwise, it will display the view query results page by default.
context
A string of "name:data" that describes the context
that generated this query result. It may be:
- "table":"<table_name>"
- "design":<design_id>
All other GET params will be passed to on_success_url (if present).
"""
# Coerce types; manage arguments
id = int(id)
query_history = authorized_get_history(request, id, must_exist=True)
# GET param: context.
context_param = request.GET.get('context', '')
# GET param: on_success_url. Default to view_results
results_url = urlresolvers.reverse(view_results,
kwargs=dict(id=str(id),
first_row=0,
last_result_len=0))
on_success_url = request.GET.get('on_success_url')
if not on_success_url:
on_success_url = results_url
# Get the server_id
server_id, state = _get_server_id_and_state(query_history)
query_history.save_state(state)
# Query finished?
if state == QueryHistory.STATE.expired:
raise PopupException(_("The result of this query has expired."))
elif state == QueryHistory.STATE.available:
return format_preserving_redirect(request, on_success_url, request.GET)
elif state == QueryHistory.STATE.failed:
# When we fetch, Beeswax server will throw us a BeeswaxException, which has the
# log we want to display.
return format_preserving_redirect(request, results_url, request.GET)
# Still running
log = db_utils.db_client(
query_history.get_query_server()).get_log(server_id)
# Keep waiting
# - Translate context into something more meaningful (type, data)
context = _parse_query_context(context_param)
return render(
'watch_wait.mako', request, {
'query': query_history,
'fwd_params': request.GET.urlencode(),
'log': log,
'hadoop_jobs': _parse_out_hadoop_jobs(log),
'query_context': context,
})
error_message, log = expand_exception(ex)
# Handle errors
if fetch_error:
return render(
'watch_results.mako', request, {
'query': query_history,
'error': True,
'error_message': error_message,
'log': log,
'hadoop_jobs': _parse_out_hadoop_jobs(log),
'query_context': context,
'can_save': False,
})
log = db_utils.db_client(query_history.get_query_server()).get_log(
query_history.server_id)
download_urls = {}
if downloadable:
for format in common.DL_FORMATS:
download_urls[format] = urlresolvers.reverse(download,
kwargs=dict(
id=str(id),
format=format))
save_form = SaveResultsForm()
has_more = True
last_result_len = long(last_result_len)
if (last_result_len != 0 and len(results.data) != last_result_len) or len(
results.data) == 0:
has_more = False
# Display the results
def save_results(request, id):
"""
Save the results of a query to an HDFS directory
"""
id = int(id)
query_history = models.QueryHistory.objects.get(id=id)
if query_history.owner != request.user:
raise PopupException(_('This action is only available to the user who submitted the query.'))
server_id, state = _get_server_id_and_state(query_history)
query_history.save_state(state)
error_msg, log = None, None
if request.method == 'POST':
# Make sure the result is available.
# Note that we may still hit errors during the actual save
if state != models.QueryHistory.STATE.available:
if state in (models.QueryHistory.STATE.failed, models.QueryHistory.STATE.expired):
msg = _('This query has %(state)s. Results unavailable.') % {'state': state}
else:
msg = _('The result of this query is not available yet.')
raise PopupException(msg)
form = beeswax.forms.SaveResultsForm(request.POST)
# Cancel goes back to results
if request.POST.get('cancel'):
return format_preserving_redirect(request, '/beeswax/watch/%s' % (id,))
if form.is_valid():
# Do save
# 1. Get the results metadata
assert request.POST.get('save')
handle = QueryHandle(id=query_history.server_id, log_context=query_history.log_context)
try:
result_meta = db_utils.db_client().get_results_metadata(handle)
except QueryNotFoundException, ex:
LOG.exception(ex)
raise PopupException(_('Cannot find query.'))
if result_meta.table_dir:
result_meta.table_dir = request.fs.urlsplit(result_meta.table_dir)[2]
# 2. Check for partitioned tables
if result_meta.table_dir is None:
raise PopupException(_('Saving results from a partitioned table is not supported. You may copy from the HDFS location manually.'))
# 3. Actual saving of results
try:
if form.cleaned_data['save_target'] == form.SAVE_TYPE_DIR:
# To dir
if result_meta.in_tablename:
raise PopupException(_('Saving results from a table to a directory is not supported. You may copy from the HDFS location manually.'))
target_dir = form.cleaned_data['target_dir']
request.fs.rename_star(result_meta.table_dir, target_dir)
LOG.debug("Moved results from %s to %s" % (result_meta.table_dir, target_dir))
query_history.save_state(models.QueryHistory.STATE.expired)
return HttpResponse(urlresolvers.reverse('filebrowser.views.view', kwargs={'path': target_dir}))
elif form.cleaned_data['save_target'] == form.SAVE_TYPE_TBL:
# To new table
try:
return _save_results_ctas(request,
query_history,
form.cleaned_data['target_table'],
result_meta)
except BeeswaxException, bex:
LOG.exception(bex)
error_msg, log = expand_exception(bex)
except WebHdfsException, ex:
raise PopupException(_('The table could not be saved.'), detail=ex)
except IOError, ex:
LOG.exception(ex)
error_msg = str(ex)
