rapids/src/data/unify_utils.R

200 lines
11 KiB
R
Raw Normal View History

library("dplyr", warn.conflicts = F)
library(stringr)
unify_ios_screen <- function(ios_screen){
# In Android we only process UNLOCK to OFF episodes. In iOS we only process UNLOCK to LOCKED episodes,
# thus, we replace LOCKED with OFF episodes (2 to 0) so we can use Android's code for iOS
ios_screen <- ios_screen %>%
# only keep consecutive pairs of 3,2 events
filter( (screen_status == 3 & lead(screen_status) == 2) | (screen_status == 2 & lag(screen_status) == 3) ) %>%
mutate(screen_status = replace(screen_status, screen_status == 2, 0))
return(ios_screen)
}
unify_ios_battery <- function(ios_battery){
# We only need to unify battery data for iOS client V1. V2 does it out-of-the-box
# V1 will not have rows where battery_status is equal to 4
if(nrow(ios_battery %>% filter(battery_status == 4)) == 0)
ios_battery <- ios_battery %>%
mutate(battery_status = replace(battery_status, battery_status == 3, 5),
battery_status = replace(battery_status, battery_status == 1, 3))
return(ios_battery)
}
unify_ios_calls <- function(ios_calls){
# Androids call types 1=incoming, 2=outgoing, 3=missed
# iOS' call status 1=incoming, 2=connected, 3=dialing, 4=disconnected
# iOS' call types based on call status: (1,2,4)=incoming=1, (3,2,4)=outgoing=2, (1,4) or (3,4)=missed=3
# Sometimes (due to a possible bug in Aware) sequences get logged on the exact same timestamp, thus 3-item sequences can be 2,3,4 or 3,2,4
# Even tho iOS stores the duration of ringing/dialing for missed calls, we set it to 0 to match Android
ios_calls <- ios_calls %>%
arrange(trace, timestamp, call_type) %>%
group_by(trace) %>%
# search for the disconnect event, as it is common to outgoing, received and missed calls
mutate(completed_call = ifelse(call_type == 4, 2, 0),
# assign the same ID to all events before a 4
completed_call = cumsum(c(1, head(completed_call, -1) != tail(completed_call, -1))),
# hack to match ID of last event (4) to that of the previous rows
completed_call = ifelse(call_type == 4, completed_call - 1, completed_call))
# We check utc_date_time and local_date_time exist because sometimes we call this function from
# download_dataset to unify multi-platform participants. At that point such time columns are missing
if("utc_date_time" %in% colnames(ios_calls) && "local_date_time" %in% colnames(ios_calls)){
ios_calls <- ios_calls %>% summarise(call_type_sequence = paste(call_type, collapse = ","), # collapse all events before a 4
# sanity check, timestamp_diff should be equal or close to duration sum
# timestamp_diff = trunc((last(timestamp) - first(timestamp)) / 1000)
# use call_duration = last(call_duration) if you want duration from pick up to hang up
# use call_duration = sum(call_duration) if you want duration from dialing/ringing to hang up
call_duration = last(call_duration),
timestamp = first(timestamp),
utc_date_time = first(utc_date_time),
local_date_time = first(local_date_time),
local_date = first(local_date),
local_time = first(local_time),
local_hour = first(local_hour),
local_minute = first(local_minute),
2020-09-14 20:21:36 +02:00
local_timezone = first(local_timezone),
assigned_segments = first(assigned_segments))
}
else {
ios_calls <- ios_calls %>% summarise(call_type_sequence = paste(call_type, collapse = ","), call_duration = sum(call_duration), timestamp = first(timestamp))
}
ios_calls <- ios_calls %>% mutate(call_type = case_when(
call_type_sequence == "1,2,4" | call_type_sequence == "2,1,4" ~ 1, # incoming
call_type_sequence == "1,4" ~ 3, # missed
call_type_sequence == "3,2,4" | call_type_sequence == "2,3,4" ~ 2, # outgoing
call_type_sequence == "3,4" ~ 4, # outgoing missed, we create this temp missed state to assign a duration of 0 below
TRUE ~ -1), # other, call sequences without a disconnect (4) event are discarded
# assign a duration of 0 to incoming and outgoing missed calls
call_duration = ifelse(call_type == 3 | call_type == 4, 0, call_duration),
# get rid of the temp missed call type, set to 2 to match Android. See https://github.com/carissalow/rapids/issues/79
call_type = ifelse(call_type == 4, 2, call_type)
) %>%
# discard sequences without an event 4 (disconnect)
filter(call_type > 0) %>%
ungroup() %>%
arrange(timestamp)
return(ios_calls)
}
clean_ios_activity_column <- function(ios_gar){
ios_gar <- ios_gar %>%
mutate(activities = str_replace_all(activities, pattern = '("|\\[|\\])', replacement = ""))
existent_multiple_activities <- ios_gar %>%
filter(str_detect(activities, ",")) %>%
group_by(activities) %>%
summarise(mutiple_activities = unique(activities)) %>%
pull(mutiple_activities)
known_multiple_activities <- c("stationary,automotive")
unkown_multiple_actvities <- setdiff(existent_multiple_activities, known_multiple_activities)
if(length(unkown_multiple_actvities) > 0){
stop(paste0("There are unkwown combinations of ios activities, you need to implement the decision of the ones to keep: ", unkown_multiple_actvities))
}
ios_gar <- ios_gar %>%
mutate(activities = str_replace_all(activities, pattern = "stationary,automotive", replacement = "automotive"))
return(ios_gar)
}
2020-10-19 21:07:12 +02:00
unify_ios_activity_recognition <- function(ios_gar){
# We only need to unify Google Activity Recognition data for iOS
# discard rows where activities column is blank
ios_gar <- ios_gar[-which(ios_gar$activities == ""), ]
# clean "activities" column of ios_gar
ios_gar <- clean_ios_activity_column(ios_gar)
# make it compatible with android version: generate "activity_name" and "activity_type" columns
ios_gar <- ios_gar %>%
mutate(activity_name = case_when(activities == "automotive" ~ "in_vehicle",
activities == "cycling" ~ "on_bicycle",
activities == "walking" ~ "walking",
activities == "running" ~ "running",
activities == "stationary" ~ "still"),
activity_type = case_when(activities == "automotive" ~ 0,
activities == "cycling" ~ 1,
activities == "walking" ~ 7,
activities == "running" ~ 8,
activities == "stationary" ~ 3,
activities == "unknown" ~ 4))
return(ios_gar)
}
2020-08-11 22:18:06 +02:00
unify_ios_conversation <- function(conversation){
if(nrow(conversation) > 0){
duration_check <- conversation %>%
select(double_convo_start, double_convo_end) %>%
mutate(start_is_seconds = double_convo_start <= 9999999999,
end_is_seconds = double_convo_end <= 9999999999) # Values smaller than 9999999999 are in seconds instead of milliseconds
start_end_in_seconds = sum(duration_check$start_is_seconds) + sum(duration_check$end_is_seconds)
if(start_end_in_seconds > 0) # convert seconds to milliseconds
conversation <- conversation %>% mutate(double_convo_start = double_convo_start * 1000, double_convo_end = double_convo_end * 1000)
}
return(conversation)
}
# This function is used in download_dataset.R
unify_raw_data <- function(dbEngine, sensor_table, sensor, timestamp_filter, aware_multiplatform_tables, device_ids, platforms){
# If platforms is 'multiple', fetch each device_id's platform from aware_device, otherwise, use those given by the user
if(length(platforms) == 1 && platforms == "multiple")
devices_platforms <- dbGetQuery(dbEngine, paste0("SELECT device_id,brand FROM aware_device WHERE device_id IN ('", paste0(device_ids, collapse = "','"), "')")) %>%
mutate(platform = ifelse(brand == "iPhone", "ios", "android"))
else
devices_platforms <- data.frame(device_id = device_ids, platform = platforms)
# Get existent tables in database
2020-10-19 21:07:12 +02:00
available_tables_in_db <- dbGetQuery(dbEngine, paste0("SELECT table_name FROM information_schema.tables WHERE table_schema='", dbGetInfo(dbEngine)$dbname,"'"))[[1]]
if(!any(sensor_table %in% available_tables_in_db))
stop(paste0("You requested data from these table(s) ", paste0(sensor_table, collapse=", "), " but they don't exist in your database ", dbGetInfo(dbEngine)$dbname))
# Parse the table names for activity recognition and conversation plugins because they are different between android and ios
ar_tables <- setNames(aware_multiplatform_tables[1:2], c("android", "ios"))
conversation_tables <- setNames(aware_multiplatform_tables[3:4], c("android", "ios"))
participants_sensordata <- list()
for(i in 1:nrow(devices_platforms)) {
row <- devices_platforms[i,]
device_id <- row$device_id
platform <- row$platform
# Handle special cases when tables for the same sensor have different names for Android and iOS (AR and conversation)
2020-10-19 21:07:12 +02:00
if(length(sensor_table) == 1)
table <- sensor_table
else if(all(sensor_table == ar_tables))
table <- ar_tables[[platform]]
2020-10-19 21:07:12 +02:00
else if(all(sensor_table == conversation_tables))
table <- conversation_tables[[platform]]
if(table %in% available_tables_in_db){
query <- paste0("SELECT * FROM ", table, " WHERE device_id IN ('", device_id, "')", timestamp_filter)
2020-10-19 21:07:12 +02:00
sensor_data <- unify_data(dbGetQuery(dbEngine, query), sensor, platform)
participants_sensordata <- append(participants_sensordata, list(sensor_data))
}else{
warning(paste0("Missing ", table, " table. We unified the data from ", paste0(devices_platforms$device_id, collapse = " and "), " but without records from this missing table for ", device_id))
}
}
unified_data <- bind_rows(participants_sensordata)
return(unified_data)
}
# This function is used in unify_ios_android.R and unify_raw_data function
2020-10-19 21:07:12 +02:00
unify_data <- function(sensor_data, sensor, platform){
if(sensor == "phone_calls" & platform == "ios"){
sensor_data = unify_ios_calls(sensor_data)
} else if(sensor == "phone_battery" & platform == "ios"){
sensor_data = unify_ios_battery(sensor_data)
} else if(sensor == "phone_activity_recognition" & platform == "ios"){
sensor_data = unify_ios_activity_recognition(sensor_data)
} else if(sensor == "phone_screen" & platform == "ios"){
sensor_data = unify_ios_screen(sensor_data)
} else if(sensor == "phone_conversation" & platform == "ios"){
2020-08-11 22:18:06 +02:00
sensor_data = unify_ios_conversation(sensor_data)
}
return(sensor_data)
}