Skip to content

Commit

Permalink
release v0.1.4 - implement tolerance
Browse files Browse the repository at this point in the history 
- simple day/amount tolerance implementation
- worst/best case scenarios are plotted as light shadow in the background
  • Loading branch information
Yann Büchau committed Jan 13, 2017
1 parent 549acf8 commit d96d513
Show file tree
Hide file tree
Showing 5 changed files with 175 additions and 123 deletions.
7 changes: 4 additions & 3 deletions debian/changelog
View file
Original file line number Diff line number Diff line change
@@ -1,8 +1,9 @@
simbuto (0.1.4) UNRELEASED; urgency=medium
simbuto (0.1.4) unstable; urgency=medium

* implement tolerance
* implement amount and day tolerance
* worst/best case scenarios are plotted as light shadow in the background

-- Yann Büchau <[email protected]> Wed, 11 Jan 2017 22:38:12 +0100
-- Yann Büchau <[email protected]> Fri, 13 Jan 2017 12:34:50 +0100

simbuto (0.1.3) unstable; urgency=medium

Expand Down
1 change: 1 addition & 0 deletions usr/lib/simbuto/python/simbuto/gui.py
View file
Original file line number Diff line number Diff line change
Expand Up @@ -480,6 +480,7 @@ def update_graph_from_editor(self, *args, size=None):
name = "{}.png".format(currentfile)
filename = os.path.join(config.personal_simbuto_dotfolder(),
"plots",name)
self.update_statusbar(_("updating graph..."))
success = self.signalmanager.emit_signal("create-graph-from-text",
filename=filename, # to this file
text = self.current_editor_content, # this text
Expand Down
6 changes: 2 additions & 4 deletions usr/lib/simbuto/python/simbuto/manager.py
View file
Original file line number Diff line number Diff line change
Expand Up @@ -130,10 +130,8 @@ def create_png_graph_from_text(self, text, filename,
# create the budget from text
budget_frame = R.read_budget_from_text(text = text)
# create the timeseries from the budget
# timeseries_frame = R.timeseries_from_budget(budget = budget_frame,
# start = start_date, end = end_date)
timeseries_frame = R.budget_ensemble(budget = budget_frame,
start = start_date, end = end_date, ensemble_size = 100)
timeseries_frame = R.timeseries_from_budget(budget = budget_frame,
start = start_date, end = end_date)
# plot to png
R.plot_budget_timeseries_to_png(filename=filename,
timeseries = timeseries_frame, width = width, height = height)
Expand Down
202 changes: 125 additions & 77 deletions usr/lib/simbuto/r/simbuto-functions.R
View file
Original file line number Diff line number Diff line change
Expand Up @@ -5,92 +5,154 @@
read_budget_from_text <- function(text) {
BUDGET <- read.csv2(text=text, stringsAsFactors = F, na.strings = c("NA"),comment.char="#")
BUDGET$amount <- as.numeric(BUDGET$amount)
if(!is.null(BUDGET$tolerance))
BUDGET$tolerance <- as.numeric(BUDGET$tolerance)
if(!is.null(BUDGET$tolerance_amount))
BUDGET$tolerance_amount <- as.numeric(BUDGET$tolerance_amount)
BUDGET$start <- as.Date(BUDGET$start,format="%F")
BUDGET$end <- as.Date(BUDGET$end,format="%F")
BUDGET$frequency[BUDGET$frequency == "monthly"] = "month"
BUDGET$frequency[BUDGET$frequency == "weekly"] = "week"
BUDGET$frequency[BUDGET$frequency == "yearly"] = "year"
BUDGET$frequency[BUDGET$frequency == "daily"] = "day"
return(BUDGET)
}

timeseries_from_budget <- function(
budget,
start = Sys.Date(), end = Sys.Date() + 365,
with_tolerance = FALSE,
random_tolerance = FALSE
ensemble_size = NULL
) {
# create empty frame with day series
all.days <- seq.Date(from = start, to = end, by = "days")
MONEY <- data.frame(day = all.days, amount = 0)
if(with_tolerance)
MONEY$mincase <- MONEY$maxcase <- 0

for (factnr in 1:nrow(budget)) { # loop over all facts
# start with empty series
worstcase <- bestcase <- undisturbed <- rep(0, nrow(MONEY))
# loop over all facts
for (factnr in 1:nrow(budget)) {
fact <- budget[factnr,] # current fact
# create sequence of occurence days
fact.start <- if(is.na(fact$start)){start}else{fact$start}
fact.end <- if(is.na(fact$end)){end}else{fact$end}
# cat("fact ",fact$title," occurs ",fact$frequency," from ",fact.start," to ",fact.end,"\n")

number.occurences <- NULL
if(fact$frequency == "once") {
number.occurences <- 1
interval <- NULL
} else if(fact$frequency == "monthly") {
interval <- "month"
} else if(fact$frequency == "yearly") {
interval <- "year"
} else if(fact$frequency == "weekly") {
interval <- "week"
} else if(fact$frequency == "dayly") {
interval <- "day"
} else {
interval <- fact$frequency
interval = fact$frequency
if(interval == "once") {
fact.end <- fact.start
interval = "day" # pick any interval, doesn't matter
}


# cat("from=",fact.start," to=",fact.end," by=",interval," length.out=",number.occurences,"\n")
occurences <- c()
if(fact.start < fact.end) {
if(is.numeric(number.occurences)){
occurences <- seq.Date(from = fact.start, to = fact.end, length.out = number.occurences)
} else {
occurences <- seq.Date(from = fact.start, to = fact.end, by = interval)
}
if(fact.start <= fact.end) {
occurences <- seq.Date(from = fact.start, to = fact.end, by = interval)
}

occurences_bool <- MONEY$day %in% occurences

# get the indices
indices <- na.omit(match(x = occurences, table = MONEY$day))
# cat("indices: ",indices)
undisturbed <- undisturbed + fact_amounts_series(
occurences = occurences_bool, fact = fact,with_tolerance = FALSE)
worstcase <- worstcase + fact_amounts_series(
occurences = occurences_bool, fact = fact,with_tolerance = TRUE,
worst_case = TRUE )
bestcase <- bestcase + fact_amounts_series(
occurences = occurences_bool, fact = fact,with_tolerance = TRUE,
worst_case = FALSE )
}

# generate random sequence
fact.amount = fact$amount
amounts <- fact.amount
fact.mincase <- fact.maxcase <- fact.amount
if(with_tolerance) { # only if specified
if(any(is.finite(fact$tolerance))) { # only if tolerance is given
fact.tolerance = abs(as.numeric(fact["tolerance"]))
if(random_tolerance) {
# cat("tolerance: ",fact.tolerance,"\n")
amounts <- runif(length(indices),
min = fact.amount - fact.tolerance,
max = fact.amount + fact.tolerance)
} else {
fact.mincase <- fact.amount - fact.tolerance
fact.maxcase <- fact.amount + fact.tolerance
}
# cumulate
MONEY$amount = cumsum(undisturbed)
MONEY$worstcase = cumsum(worstcase)
MONEY$bestcase = cumsum(bestcase)
# empty data frame
return(MONEY)
}

fact_amounts_series <- function(
fact, # the fact dataframe row/list
occurences, # boolean vector with TRUE where the fact occurs, output has same length
with_tolerance = FALSE, # use the tolerance?
worst_case = FALSE, # TRUE = worst_case, FALSE = best_case
random_tolerance = FALSE # if using the tolerance, randomize?
) {
stopifnot(nrow(fact)==1)

# the indices where the fact occurs
indices <- which(occurences)
# the output sequence starts with zeros everywhere
out <- rep(0,length(occurences))
# output length
N <- length(out)

# the tolerances
fact_tolerance_day <- 0
if(any(is.finite(fact$tolerance_day)))
fact_tolerance_day = as.integer(abs(fact$tolerance_day))
fact_tolerance_amount <- 0
if(any(is.finite(fact$tolerance_amount)))
fact_tolerance_amount = abs(fact$tolerance_amount)
# fact data
fact_amount <- 0
if(any(is.finite(fact$amount)))
fact_amount = fact$amount

if(with_tolerance) {
if(random_tolerance) {
# modify amount randomly
amounts <- runif( n = length(indices),
min = fact_amount - fact_tolerance_amount,
max = fact_amount + fact_tolerance_amount
)
# modify indices randomly
indices <- indices + runif( n = length(indices),
min = - fact_tolerance_amount,
max = + fact_tolerance_amount
)
} else {
if(worst_case) {
# worst case: all costs are highest
# worst case: all incomes are lowest
amounts <- rep(fact_amount - fact_tolerance_amount, length(indices))
# worst case: all costs are earliest
# worst case: all incomes are latest
indices <- indices + sign(fact_amount) * fact_tolerance_day
} else {
# best case: all costs are lowest
# best case: all incomes are highest
amounts <- rep(fact_amount + fact_tolerance_amount, length(indices))
# best case: all costs are latest
# best case: all incomes are earliest
indices <- indices - sign(fact_amount) * fact_tolerance_day
}
MONEY[indices,"mincase"] = MONEY[indices,"mincase"] + fact.mincase
MONEY[indices,"maxcase"] = MONEY[indices,"maxcase"] + fact.maxcase
# fix indices that lie outside the output vector
# cat("indices before fixing: ",indices,"\n")
# cat("amounts before fixing: ",amounts,"\n")
tooearly <- which(indices < 1)
# cat("too early indices: ",indices[tooearly],"\n")
# stopifnot(length(tooearly)==0)
toolate <- which(indices > N)
# cat("too late indices: ",indices[toolate],"\n")
outside <- sort(unique(c(tooearly,toolate)))
out[1] <- out[1] + sum(amounts[tooearly])
out[N] <- out[N] + sum(amounts[toolate])
if(length(outside)>0) {
amounts <- amounts[-(outside)]
indices <- indices[-(outside)]
}

# add to time series
MONEY[indices,"amount"] = MONEY[indices,"amount"] + amounts
# cat("indices after fixing: ",indices,"\n")
# cat("amounts after fixing: ",amounts,"\n")
}
} else {
# keep amount
amounts <- rep(fact_amount, length(indices))
}
MONEY$amount = cumsum(MONEY$amount)
MONEY$mincase = cumsum(MONEY$mincase)
MONEY$maxcase = cumsum(MONEY$maxcase)
# empty data frame
return(MONEY)
# cat("amounts before putting into out: ",amounts,"\n")

# set the amounts to the indices
out[indices] <- amounts
# cat("out: ",out,"\n")

# return out vector
return(out)
}

budget_ensemble<- function( budget,
Expand All @@ -103,7 +165,7 @@ budget_ensemble<- function( budget,
random_tolerance = FALSE)
# the ensemble out starts with the bare run
ENSEMBLE_OUT <- timeseries_without_tolerance
if(any(is.finite(budget$tolerance))) {
if(any(is.finite(budget$tolerance_amount))) {
# create ensemble matrix
ENSEMBLE <- matrix(NA,nrow=ensemble_size, ncol = nrow(ENSEMBLE_OUT))
# do the runs
Expand All @@ -128,8 +190,8 @@ budget_ensemble<- function( budget,
}

plot_budget_timeseries <- function(timeseries) {
plotrange <- range(c(timeseries$amount,timeseries$mincase,
timeseries$maxcase,timeseries$ensmin,timeseries$ensmax))
plotrange <- range(c(timeseries$amount,timeseries$worstcase,
timeseries$bestcase,timeseries$ensmin,timeseries$ensmax))
# base plot
plot(timeseries$day,timeseries$amount,type="n",xaxt="n",yaxt="n"
,ylab="",xlab="",ylim=plotrange,
Expand Down Expand Up @@ -158,23 +220,11 @@ plot_budget_timeseries <- function(timeseries) {
do.call(rect, bad)

# worst/best cases
if(!is.null(timeseries$mincase) & !is.null(timeseries$maxcase)) {
if(!is.null(timeseries$worstcase) & !is.null(timeseries$bestcase)) {
polygon(x = c(timeseries$day,rev(timeseries$day)),
y = c(timeseries$mincase,rev(timeseries$maxcase)),
y = c(timeseries$worstcase,rev(timeseries$bestcase)),
col = "#00000022",border=NA)
}
# ensemble max/min
if(!is.null(timeseries$ensmin) & !is.null(timeseries$ensmax)) {
polygon(x = c(timeseries$day,rev(timeseries$day)),
y = c(timeseries$ensmax,rev(timeseries$ensmin)),
col = "#00000033",border=NA)
}
# ensemble quantiles
if(!is.null(timeseries$ensquant25) & !is.null(timeseries$ensquant75)) {
polygon(x = c(timeseries$day,rev(timeseries$day)),
y = c(timeseries$ensquant75,rev(timeseries$ensquant25)),
col = "#00000044",border=NA)
}
# raw run
lines(x = timeseries$day, y = timeseries$amount
,lwd = 4
Expand All @@ -190,9 +240,7 @@ plot_budget_timeseries_to_png <- function(timeseries,filename,width=600,height=4

#### read data ####
# BUDGET <- read_budget_from_text(readLines("~/Downloads/budget.simbuto"))
# # MONEY <- timeseries_from_budget(budget = BUDGET, with_tolerance = TRUE)
# MONEY <- budget_ensemble(budget = BUDGET)
# MONEY <- timeseries_from_budget(budget = BUDGET)
# cat("plotting...")
# plot_budget_timeseries(MONEY)
# cat("done!\n")
#
Loading

0 comments on commit d96d513

Please sign in to comment.