xref: /aosp_15_r20/external/rappor/tests/analyze_assoc.R (revision 2abb31345f6c95944768b5222a9a5ed3fc68cc00) 
1#!/usr/bin/env Rscript
2#
3# Copyright 2015 Google Inc. All rights reserved.
4#
5# Licensed under the Apache License, Version 2.0 (the "License");
6# you may not use this file except in compliance with the License.
7# You may obtain a copy of the License at
8#
9#     http://www.apache.org/licenses/LICENSE-2.0
10#
11# Unless required by applicable law or agreed to in writing, software
12# distributed under the License is distributed on an "AS IS" BASIS,
13# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
14# See the License for the specific language governing permissions and
15# limitations under the License.
16
17# Reads map files, report files, and RAPPOR parameters to run
18# an EM algorithm to estimate joint distribution over two or more variables
19#
20# Usage:
21#       $ ./analyze_assoc.R -map1 map_1.csv -map2 map_2.csv \
22#                                 -reports reports.csv \
23# Inputs: map1, map2, reports, params
24#         see how options are parsed below for more information
25# Outputs:
26#         prints a table with estimated joint probability masses
27#         over candidate strings
28#         Ex.
29#                 ssl   nossl
30#         intel   0.1   0.3
31#         google  0.5   0.1
32
33library("optparse")
34
35options(stringsAsFactors = FALSE)
36
37if(!interactive()) {
38  option_list <- list(
39    # Flags
40    make_option(c("--map1", "-m1"), default = "map_1.csv",
41                help = "Hashed candidates for 1st variable"),
42    make_option(c("--map2", "-m2"), default = "map_2.csv",
43                help = "Hashed candidates for 2nd variable"),
44    make_option(c("--reports", "-r"), default = "reports.csv",
45                help = "File with raw reports as <cohort, report1, report2>"),
46    make_option(c("--params", "-p"), default = "params.csv",
47                help = "Filename for RAPPOR parameters")
48  )
49  opts <- parse_args(OptionParser(option_list = option_list))
50}
51
52source("../analysis/R/encode.R")
53source("../analysis/R/decode.R")
54source("../analysis/R/simulation.R")
55source("../analysis/R/read_input.R")
56source("../analysis/R/association.R")
57
58# This function processes the maps loaded using ReadMapFile
59# Association analysis requires a map object with a map
60# field that has the map split into cohorts and an rmap field
61# that has all the cohorts combined
62# Arguments:
63#       map = map object with cohorts as sparse matrix in
64#             object map$map
65#             This is the expected object from ReadMapFile
66#       params = data field with parameters
67# TODO(pseudorandom): move this functionality to ReadMapFile
68ProcessMap <- function(map, params) {
69  map$rmap <- map$map
70  split_map <- function(i, map_struct) {
71    numbits <- params$k
72    indices <- which(as.matrix(
73      map_struct[((i - 1) * numbits + 1):(i * numbits),]) == TRUE,
74      arr.ind = TRUE)
75    sparseMatrix(indices[, "row"], indices[, "col"],
76                 dims = c(numbits, max(indices[, "col"])))
77  }
78  map$map <- lapply(1:params$m, function(i) split_map(i, map$rmap))
79  map
80}
81
82main <- function(opts) {
83  ptm <- proc.time()
84
85  params <- ReadParameterFile(opts$params)
86  opts_map <- list(opts$map1, opts$map2)
87  map <- lapply(opts_map, function(o)
88                  ProcessMap(ReadMapFile(o, params = params),
89                             params = params))
90  # Reports must be of the format
91  #     cohort no, rappor bitstring 1, rappor bitstring 2
92  reportsObj <- read.csv(opts$reports,
93                         colClasses = c("integer", "character", "character"),
94                         header = FALSE)
95
96  # Parsing reportsObj
97  # ComputeDistributionEM allows for different sets of cohorts
98  # for each variable. Here, both sets of cohorts are identical
99  co <- as.list(reportsObj[1])[[1]]
100  cohorts <- list(co, co)
101  # Parse reports from reportObj cols 2 and 3
102  reports <- lapply(1:2, function(x) as.list(reportsObj[x + 1]))
103
104  # Split strings into bit arrays (as required by assoc analysis)
105  reports <- lapply(1:2, function(i) {
106    # apply the following function to each of reports[[1]] and reports[[2]]
107    lapply(reports[[i]][[1]], function(x) {
108      # function splits strings and converts them to numeric values
109      as.numeric(strsplit(x, split = "")[[1]])
110    })
111  })
112
113  joint_dist <- ComputeDistributionEM(reports, cohorts, map,
114                                      ignore_other = TRUE,
115                                      params, marginals = NULL,
116                                      estimate_var = FALSE)
117  # TODO(pseudorandom): Export the results to a file for further analysis
118  print("JOINT_DIST$FIT")
119  print(joint_dist$fit)
120  print("PROC.TIME")
121  print(proc.time() - ptm)
122}
123
124if(!interactive()) {
125  main(opts)
126}