Modify db generation script to find same/better replacements for all items
This commit is contained in:
@@ -28,6 +28,8 @@ sys.path.insert(0, os.path.realpath(os.path.join(path, '..')))
|
||||
|
||||
import json
|
||||
import argparse
|
||||
import itertools
|
||||
|
||||
|
||||
CATEGORIES_TO_REMOVE = [
|
||||
30 # Apparel
|
||||
@@ -174,6 +176,109 @@ def main(db, json_path):
|
||||
newData.append(newRow)
|
||||
return newData
|
||||
|
||||
def fillReplacements(tables):
|
||||
|
||||
def compareAttrs(attrs1, attrs2, attrHig):
|
||||
"""
|
||||
Compares received attribute sets. Returns:
|
||||
- 0 if sets are different
|
||||
- 1 if sets are exactly the same
|
||||
- 2 if first set is strictly better
|
||||
- 3 if second set is strictly better
|
||||
"""
|
||||
if set(attrs1) != set(attrs2):
|
||||
return 0
|
||||
if all(attrs1[aid] == attrs2[aid] for aid in attrs1):
|
||||
return 1
|
||||
if all(
|
||||
(attrs1[aid] >= attrs2[aid] and attrHig[aid]) or
|
||||
(attrs1[aid] <= attrs2[aid] and not attrHig[aid])
|
||||
for aid in attrs1
|
||||
):
|
||||
return 2
|
||||
if all(
|
||||
(attrs2[aid] >= attrs1[aid] and attrHig[aid]) or
|
||||
(attrs2[aid] <= attrs1[aid] and not attrHig[aid])
|
||||
for aid in attrs1
|
||||
):
|
||||
return 3
|
||||
return 0
|
||||
|
||||
skillReqAttribs = {
|
||||
182: 277,
|
||||
183: 278,
|
||||
184: 279,
|
||||
1285: 1286,
|
||||
1289: 1287,
|
||||
1290: 1288}
|
||||
skillReqAttribsFlat = set(skillReqAttribs.keys()).union(skillReqAttribs.values())
|
||||
# Get data on type groups
|
||||
typesGroups = {}
|
||||
for row in tables['evetypes']:
|
||||
typesGroups[row['typeID']] = row['groupID']
|
||||
# Get data on type attributes
|
||||
typesNormalAttribs = {}
|
||||
typesSkillAttribs = {}
|
||||
for row in tables['dgmtypeattribs']:
|
||||
# TODO: add attribute ignore stuff (like meta level) here
|
||||
typesAttribs = typesSkillAttribs if row['attributeID'] in skillReqAttribsFlat else typesNormalAttribs
|
||||
typeAttribs = typesAttribs.setdefault(row['typeID'], {})
|
||||
typeAttribs[row['attributeID']] = row['value']
|
||||
# Get data on skill requirements
|
||||
typesSkillReqs = {}
|
||||
for typeID, typeAttribs in typesSkillAttribs.items():
|
||||
typeSkillAttribs = typesSkillAttribs.get(typeID, {})
|
||||
if not typeSkillAttribs:
|
||||
continue
|
||||
typeSkillReqs = typesSkillReqs.setdefault(typeID, {})
|
||||
for skillreqTypeAttr, skillreqLevelAttr in skillReqAttribs.items():
|
||||
try:
|
||||
skillType = int(typeSkillAttribs[skillreqTypeAttr])
|
||||
skillLevel = int(typeSkillAttribs[skillreqLevelAttr])
|
||||
except (KeyError, ValueError):
|
||||
continue
|
||||
typeSkillReqs[skillType] = skillLevel
|
||||
# Get data on attribute highIsGood flag
|
||||
attrHig = {}
|
||||
for row in tables['dgmattribs']:
|
||||
attrHig[row['attributeID']] = bool(row['highIsGood'])
|
||||
# As EVE affects various types mostly depending on their group or skill requirements,
|
||||
# we're going to group various types up this way
|
||||
groupedData = {}
|
||||
for row in tables['evetypes']:
|
||||
typeID = row['typeID']
|
||||
typeAttribs = typesNormalAttribs.get(typeID, {})
|
||||
# Ignore stuff w/o attributes
|
||||
if not typeAttribs:
|
||||
continue
|
||||
# We need only skill types, not levels for keys
|
||||
typeSkillreqs = frozenset(typesSkillReqs.get(typeID, {}))
|
||||
typeGroup = typesGroups[typeID]
|
||||
groupData = groupedData.setdefault((typeGroup, typeSkillreqs), [])
|
||||
groupData.append((typeID, typeAttribs))
|
||||
same = {}
|
||||
better = {}
|
||||
# Now, go through composed groups and for every item within it find items which are
|
||||
# the same and which are better
|
||||
for groupData in groupedData.values():
|
||||
for type1, type2 in itertools.combinations(groupData, 2):
|
||||
comparisonResult = compareAttrs(type1[1], type2[1], attrHig)
|
||||
# Equal
|
||||
if comparisonResult == 1:
|
||||
same.setdefault(type1[0], set()).add(type2[0])
|
||||
same.setdefault(type2[0], set()).add(type1[0])
|
||||
# First is better
|
||||
elif comparisonResult == 2:
|
||||
better.setdefault(type2[0], set()).add(type1[0])
|
||||
# Second is better
|
||||
elif comparisonResult == 3:
|
||||
better.setdefault(type1[0], set()).add(type2[0])
|
||||
# Put this data into types table so that normal process hooks it up
|
||||
for row in tables['evetypes']:
|
||||
typeID = row['typeID']
|
||||
row['replaceSame'] = ','.join('{}'.format(tid) for tid in sorted(same.get(typeID, ())))
|
||||
row['replaceBetter'] = ','.join('{}'.format(tid) for tid in sorted(better.get(typeID, ())))
|
||||
|
||||
data = {}
|
||||
|
||||
# Dump all data to memory so we can easely cross check ignored rows
|
||||
@@ -190,6 +295,8 @@ def main(db, json_path):
|
||||
tableData = convertClones(tableData)
|
||||
data[jsonName] = tableData
|
||||
|
||||
fillReplacements(data)
|
||||
|
||||
# Set with typeIDs which we will have in our database
|
||||
# Sometimes CCP unpublishes some items we want to have published, we
|
||||
# can do it here - just add them to initial set
|
||||
|
||||
Reference in New Issue
Block a user