report.Rmd

## Build Times

```{r preliminaries, include=FALSE}
library(ggplot2)
library(plyr)

options(width=120)
options(scipen=999)
```

```{r load, include=FALSE}
# Load data frame and prep.
df <- read.csv("times.csv", header=FALSE)
colnames(df) <- c("timestamp", "order", "task", "success", "did_work", "skipped", "milliseconds")

# Correct mis-factoring of task and boolean fields
df$task <- as.character(df$task)
df$success <- as.logical(as.character(df$success))
df$did_work <- as.logical(as.character(df$did_work))
df$skipped <- as.logical(as.character(df$skipped))

# Create date objects from timestamp field for display
df$date <- as.POSIXct(df$timestamp/1000, origin="1970-01-01")

# Add seconds fields
df$seconds <- floor(df$milliseconds / 1000)

# Sort by timestamp and order
df <- df[with(df, order(timestamp, order)),]
row.names(df) <- 1:nrow(df)
```


```{r echo=FALSE}
total <- ddply(df, .(date), summarise, milliseconds = sum(milliseconds))
total$seconds = total$milliseconds/1000
total$minutes = total$milliseconds/1000/60
total.ecdf <- ecdf(total$milliseconds)

df.required_work <- df[!df$skipped & df$did_work,]
```

Data collected on `r length(unique(df$timestamp))` Gradle builds running for
`r format(sum(df$milliseconds)/1000/60/60, digits=1)` hours in total.

* The median build time was `r format(median(total$seconds), digits=1)` seconds.
* `r format(100 - total.ecdf(10*1000) * 100, digits=1)`% of builds took longer
  than ten seconds to complete.
* `r format(100 - total.ecdf(60*1000) * 100, digits=1)`% of builds took longer
  than one minute to complete.
* `r format(100 - length(df.required_work$timestamp)/length(df$timestamp)*100, digits=1)`%
  of subtasks were skipped or required no work.


```{r echo=FALSE, fig.width=10, fig.height=7}
ggplot(total, aes(x = date, y = minutes)) +
  geom_bar(stat="identity", colour = "black") +
  theme(axis.title.x = element_blank()) + 
  ylab("Build Time (minutes)") +
  ggtitle("Build History") +
  theme(legend.title = element_blank())
```

```{r echo=FALSE, fig.width=10, fig.height=7}
ggplot(total, aes(x = minutes)) +
  geom_histogram(binwidth = .25, colour = "black") +
  xlab("Build Time (minutes)") +
  theme(axis.title.y = element_blank()) + 
  ggtitle("Build Time Distribution") +
  theme(legend.title = element_blank())
```

The slowest ten subtargets requiring work ordered by aggregate total time were:

```{r echo=FALSE, comment=NA}
task <- ddply(df.required_work, .(task), summarise, 
              executions = length(date), 
              median_seconds = median(seconds), 
              sd_seconds = sd(seconds), 
              max_seconds = max(seconds),
              total_seconds = sum(seconds),
              mean_seconds = mean(seconds))

# Sort by total time
task.total_seconds <- task[with(task, order(-total_seconds)),]
row.names(task.total_seconds) <- 1:nrow(task.total_seconds)

# Display top ten by median build time
task.total_seconds[1:10,c('task', 'executions', 'total_seconds', 'median_seconds')]
```

The slowest ten subtargets requiring work ordered by median time were:

```{r echo=FALSE, comment=NA}
# Sort by median time
task.median_seconds <- task[with(task, order(-median_seconds)),]
row.names(task.median_seconds) <- 1:nrow(task.median_seconds)

# Display top ten by median build time
task.median_seconds[1:10,c('task', 'executions', 'median_seconds')]
```