Skip to content

Commit

Permalink
minor fixes
Browse files Browse the repository at this point in the history
  • Loading branch information
@thelostone-mc
thelostone-mc committed Dec 16, 2019
1 parent 3dd88fa commit 3edef90
Showing 1 changed file with 52 additions and 78 deletions.
130 changes: 52 additions & 78 deletions app/grants/clr.py
View file
Original file line number Diff line number Diff line change
Expand Up @@ -29,12 +29,14 @@
from grants.models import Contribution, Grant, PhantomFunding
from perftools.models import JSONStore

CLR_DISTRIBUTION_AMOUNT = 100000
CLR_DISTRIBUTION_AMOUNT = 100000.0
LOWER_THRESHOLD = 0.0
CLR_START_DATE = dt.datetime(2019, 9, 15, 0, 0)


'''
Helper function that translates existing grant data structure to a list of lists.
Helper function that translates existing grant data structure
to a list of lists.
Args:
{
Expand All @@ -60,7 +62,9 @@ def translate_data(grants_data):


'''
Helper function that aggregates contributions by contributor, and then uses the aggregated contributors by contributor and calculates total contributions by unique pairs.
Helper function that aggregates contributions by contributor, and then
uses the aggregated contributors by contributor and calculates total
contributions by unique pairs.
Args:
from translate_data:
Expand Down Expand Up @@ -93,124 +97,89 @@ def aggregate_contributions(grant_contributions):


'''
Helper function that runs the pairwise clr formula while "binary" searching for the correct threshold.
Helper function that runs the pairwise clr formula while "binary"
searching for the correct threshold.
Args:
set variables:
lower_bound: set at 0.0
total_pot: set at 100000.0
from the helper function aggregate_contributions:
aggregated_contributions: {grant_id (str): {user_id (str): aggregated_amount (float)}}
pair_totals: {user_id (str): {user_id (str): pair_total (float)}}
aggregated_contributions : {
grant_id (str): {
user_id (str): aggregated_amount (float)
}
}
pair_totals : { user_id (str): { user_id (str): pair_total (float) } }
total_pot : (float)
lower_bound : (float)
Returns:
bigtot: should equal total pot
totals: clr totals
bigtot: (float)
totals: [
{
id: (str),
clr_amount: (float)
}
]
'''
def iter_threshold(aggregated_contributions, pair_totals, lower_bound=0.0, total_pot=100000.0):
def iter_threshold(aggregated_contributions, pair_totals, total_pot, lower_bound):
lower = lower_bound
upper = total_pot
iterations = 0

while iterations < 100:
threshold = (lower + upper) / 2
iterations += 1
if iterations == 100:
print("--- %s seconds ---" % (time.time() - start_time))
print(f'iterations reached, bigtot at {bigtot}')
break
break # break at 100th iteration
bigtot = 0
totals = []
# single donation doesn't get a match

for proj, contribz in aggregated_contributions.items():
tot = 0
for k1, v1 in contribz.items():
for k2, v2 in contribz.items():
if k2 > k1: # remove pairs
# # pairwise matching formula
# tot += (v1 * v2) ** 0.5 * min(1, threshold / pair_totals[k1][k2])
# vitalik's division formula
if k2 > k1: # ensure (k1,k2) and (k2,k1) are counted only once
tot += ((v1 * v2) ** 0.5) / (pair_totals[k1][k2] / threshold + 1)
bigtot += tot
# totals.append((proj, tot))
totals.append({'id': proj, 'clr_amount': tot})
# print(f'threshold {threshold} yields bigtot {bigtot} vs totalpot {total_pot} at iteration {iterations}')
if bigtot == total_pot:
print("--- %s seconds ---" % (time.time() - start_time))
print(f'bigtot {bigtot} = total_pot {total_pot} with threshold {threshold}')
print(totals)
break
elif bigtot < total_pot:
lower = threshold
elif bigtot > total_pot:
upper = threshold
return bigtot, totals
return bigtot, totals


'''
Clubbed function that intakes grant data, calculates necessary intermediate calculations, and spits out clr calculations.
Clubbed function that intakes grant data, calculates necessary
intermediate calculations, and spits out clr calculations.
Args:
{
grant_contributions: {
'id': (string) ,
'contibutions' : [
'contributions' : [
{
contributor_profile (str) : contribution_amount (int)
}
]
}
total_pot (float)
lower_bound (float)
Returns:
bigtot: should equal total pot
totals: clr totals
'''
def calculate_clr(grants_data):
grants_list = translate_data(grants_data)
def grants_clr_calculate (grant_contributions, total_pot, lower_bound):
print(f'grant_contributions {grant_contributions}')
grants_list = translate_data(grant_contributions)
print(f'grants_list {grants_list}')
aggregated_contributions, pair_totals = aggregate_contributions(grants_list)
bigtot, totals = iter_threshold(aggregated_contributions, pair_totals)
print(f'aggregated_contributions {aggregated_contributions}, pair_totals{pair_totals}')
bigtot, totals = iter_threshold(aggregated_contributions, pair_totals, total_pot, lower_bound)
return bigtot, totals


'''
Given the total pot and grants and it's contirbutions,
it uses binary search to find out the threshold so
that the entire pot can be distributed based on it's contributions
Args:
total_pot: (int),
grant_contributions: object,
min_threshold: (int)
max_threshold: (int)
iterations: (int)
previous_threshold: (int)
Returns:
grants_clr (object)
total_clr (int)
threshold (int)
iterations (int)
'''
def grants_clr_calculate (total_pot, grant_contributions, min_threshold, max_threshold, iterations = 0, previous_threshold=None):
if len(grant_contributions) == 0:
return 0, 0, 0, 0

iterations += 1
threshold = (max_threshold + min_threshold) / 2
total_clr, grants_clrs = calculate_clr(grant_contributions)

if iterations == 200 or total_pot == threshold or previous_threshold == threshold:
# No more accuracy to be had
return grants_clrs, total_clr, threshold, iterations
if total_clr > total_pot:
max_threshold = threshold
elif total_clr < total_pot:
min_threshold = threshold
else:
return grants_clrs, total_clr, threshold, iterations

return grants_clr_calculate(total_pot, grant_contributions, min_threshold, max_threshold, iterations, threshold)

def generate_random_contribution_data():
import random
contrib_data = []
Expand All @@ -223,8 +192,14 @@ def generate_random_contribution_data():
number_of_profiles = 17

for grant_id in range(grants_to_use):
contrib_data.append({'id': grant_id,
'contributions': [{str(profile_id): random.randint(low_donation, high_donation)} for profile_id in range(random.randint(1, number_of_profiles))]})
contrib_data.append({
'id': grant_id,
'contributions': [
{
str(profile_id): random.randint(low_donation, high_donation)
} for profile_id in range(random.randint(1, number_of_profiles))
]
})
return contrib_data


Expand All @@ -237,7 +212,7 @@ def calculate_clr_for_donation(donation_grant, donation_amount, total_pot, base_
# add this donation with a new profile (id 99999999999) to get impact
grant_contribution['contributions'].append({'999999999999': donation_amount})

grants_clr, _, _, _ = grants_clr_calculate(CLR_DISTRIBUTION_AMOUNT, grant_contributions, 0, CLR_DISTRIBUTION_AMOUNT)
_, grants_clr = grants_clr_calculate(grant_contributions, total_pot, LOWER_THRESHOLD)

# find grant we added the contribution to and get the new clr amount
for grant_clr in grants_clr:
Expand Down Expand Up @@ -294,8 +269,7 @@ def predict_clr(random_data=False, save_to_db=False, from_date=None, clr_type=No
# use random contribution data for testing
contrib_data = generate_random_contribution_data()

#print('\n\ncontributions data:')
#print(contrib_data)
#print(f'\n contributions data: {contrib_data} \n')

# calculate clr given additional donations
for grant in grants:
Expand Down

0 comments on commit 3edef90

Please sign in to comment.