Makes CmdFind querying use iterators to minimize memory ballooning. Refactors some redundant rendering code.

This commit is contained in:
Johnny 2020-01-10 20:56:44 +00:00
parent 716a0b20e3
commit 5339b80743
2 changed files with 73 additions and 39 deletions

View file

@ -3,7 +3,7 @@ Building and world design commands
"""
import re
from django.conf import settings
from django.db.models import Q
from django.db.models import Q, Min, Max
from evennia.objects.models import ObjectDB
from evennia.locks.lockhandler import LockException
from evennia.commands.cmdhandler import get_and_merge_cmdsets
@ -2778,8 +2778,35 @@ class CmdFind(COMMAND_DEFAULT_CLASS):
switches.append("loc")
searchstring = self.lhs
low, high = 1, ObjectDB.objects.all().order_by("-id")[0].id
try:
# Try grabbing the actual min/max id values by database aggregation
qs = ObjectDB.objects.values('id').aggregate(low=Min('id'), high=Max('id'))
low, high = sorted(qs.values())
if not (low and high):
raise ValueError(f"{self.__class__.__name__}: Min and max ID not returned by aggregation; falling back to queryset slicing.")
except Exception as e:
logger.log_trace(e)
# If that doesn't work for some reason (empty DB?), guess the lower
# bound and do a less-efficient query to find the upper.
low, high = 1, ObjectDB.objects.all().order_by("-id")[0].id
if self.rhs:
# Check that rhs is either a valid dbref or dbref range
try:
# Get rid of # signs, split on hyphen or space and cast all to int.
# Then sort by number to get the lowest and highest values
# comprising the bounds.
bounds = sorted(int(x) for x in re.split('[-\s]+', self.rhs.strip().replace('#', '')))
except ValueError:
caller.msg("Invalid dbref range provided (not a number).")
return
low = bounds[0]
if len(bounds) > 1:
high = bounds[-1]
"""
if "-" in self.rhs:
# also support low-high syntax
limlist = [part.lstrip("#").strip() for part in self.rhs.split("-", 1)]
@ -2790,6 +2817,7 @@ class CmdFind(COMMAND_DEFAULT_CLASS):
low = max(low, int(limlist[0]))
if len(limlist) > 1 and limlist[1].isdigit():
high = min(high, int(limlist[1]))
"""
low = min(low, high)
high = max(low, high)
@ -2868,51 +2896,45 @@ class CmdFind(COMMAND_DEFAULT_CLASS):
results = ObjectDB.objects.filter(keyquery | aliasquery).distinct()
nresults = results.count()
# Use iterator to minimize memory ballooning on large result sets
results = results.iterator()
if nresults:
# convert result to typeclasses.
results = [result for result in results]
# filter results by typeclasses, if requested
obj_ids = []
if "room" in switches:
results = [
obj for obj in results if inherits_from(obj, ROOM_TYPECLASS)
]
obj_ids.extend([
obj.id for obj in results if inherits_from(obj, ROOM_TYPECLASS)
])
if "exit" in switches:
results = [
obj for obj in results if inherits_from(obj, EXIT_TYPECLASS)
]
obj_ids.extend([
obj.id for obj in results if inherits_from(obj, EXIT_TYPECLASS)
])
if "char" in switches:
results = [
obj for obj in results if inherits_from(obj, CHAR_TYPECLASS)
]
nresults = len(results)
obj_ids.extend([
obj.id for obj in results if inherits_from(obj, CHAR_TYPECLASS)
])
if obj_ids:
filtered_result_qs = ObjectDB.objects.filter(id__in=set(obj_ids)).distinct()
nresults = filtered_result_qs.count()
# Keep using iterator to minimize memory ballooning
results = filtered_result_qs.iterator()
# still results after type filtering?
if nresults:
if nresults > 1:
string = "|w%i Matches|n(#%i-#%i%s):" % (
nresults,
low,
high,
restrictions,
)
for res in results:
string += "\n |g%s - %s|n" % (
res.get_display_name(caller),
res.path,
)
else:
string = "|wOne Match|n(#%i-#%i%s):" % (low, high, restrictions)
string += "\n |g%s - %s|n" % (
results[0].get_display_name(caller),
results[0].path,
)
if "loc" in self.switches and nresults == 1 and results[0].location:
string += " (|wlocation|n: |g{}|n)".format(
results[0].location.get_display_name(caller)
)
if nresults > 1: header = f'{nresults} Matches'
else: header = 'One Match'
string = f"|w{header}|n(#{low}-#{high}{restrictions}):"
for res in results:
string += f"\n |g{res.get_display_name(caller)} - {res.path}|n"
if "loc" in self.switches and nresults == 1 and res and res.location:
string += f" (|wlocation|n: |g{res.location.get_display_name(caller)}|n)"
else:
string = "|wMatch|n(#%i-#%i%s):" % (low, high, restrictions)
string += "\n |RNo matches found for '%s'|n" % searchstring
string = "|wMatch|n(#{low}-#{high}{restrictions}):"
string += "\n |RNo matches found for '{searchstring}'|n"
# send result
caller.msg(string.strip())

View file

@ -1214,6 +1214,18 @@ class TestBuilding(CommandTest):
self.call(building.CmdFind(), "/room Obj")
self.call(building.CmdFind(), "/exit Obj")
self.call(building.CmdFind(), "/exact Obj", "One Match")
# Test bogus dbref range with no search term
self.call(building.CmdFind(), "= obj", "Invalid dbref range provided (not a number).")
self.call(building.CmdFind(), "= #1a", "Invalid dbref range provided (not a number).")
# Test valid dbref ranges with no search term
self.call(building.CmdFind(), "=#1", "7 Matches(#1-#7)")
self.call(building.CmdFind(), "=1-2", "2 Matches(#1-#2):")
self.call(building.CmdFind(), "=1 - 2", "2 Matches(#1-#2):")
self.call(building.CmdFind(), "=1- #2", "2 Matches(#1-#2):")
self.call(building.CmdFind(), "=1-#2", "2 Matches(#1-#2):")
self.call(building.CmdFind(), "=#1-2", "2 Matches(#1-#2):")
def test_script(self):
self.call(building.CmdScript(), "Obj = ", "No scripts defined on Obj")