tti-analysis.Rmd

---
title: "TTI Analysis"
author: "tdresser@chromium.org"
date: "December 12, 2016"
output: html_document
---

```{r setup, include=FALSE}
knitr::opts_chunk$set(echo = FALSE)
library(ggplot2)
library(plotly)
library(reshape2)
library(GGally)
```

# Comparison of Time To Interactive Definitions.

```{r}
df = read.csv("data.csv")
df[12:13] <- list(NULL)
df$ttifmp = as.double(df$firstInteractive.FMP) / 1000.0
dfm <- reshape2:::melt.data.frame(df, id=c("url", "trace", "ttifmp"), 
            variable.name = "measurement",
            value.name = "ms")

dfm$ms <- suppressWarnings(as.double(dfm$ms)) # Ignore coercion warnings.
dfm$seconds <- dfm$ms / 1000.0
means = aggregate(ms ~ measurement, data = dfm, FUN = mean)
ordered = means[order(means$ms),][['measurement']]
dfm$measurement = factor(dfm$measurement, levels = ordered)
dfm <- na.omit(dfm)
```

```{r graph, fig.width=10, fig.height=12}
# Breaks used for log graph. 
b = c(5,7.5,10,15,20,30,40,50,60,80,100,120)

jitter_plot <- function(df) {
  p <- ggplot(aes(x=measurement, y=seconds, color=ttifmp), data=df) + 
     geom_jitter(alpha = 0.3, aes(text=url)) + 
     scale_y_continuous(
         trans = "log10", 
         breaks = b,
         labels = b) + 
     theme(
         axis.text.x = element_text(angle = 90, hjust = 1, size=15),
         axis.text.y = element_text(size=15),
         panel.grid.minor.y = element_blank()) +
     scale_color_distiller(palette = "Spectral") +
     theme(axis.title.x = element_blank())
  return(p)
}

p <- jitter_plot(dfm)
print(p)
```

## Interactive version of above graph.
Low number of samples, to make the graph lighter.

```{r graph_interactive, fig.width=10, fig.height=12}
sample_ratio <- 0.1
dfm_samples <- dfm[sample(nrow(dfm), round(nrow(dfm) * sample_ratio)),]
p <- jitter_plot(dfm_samples)
ggplotly(tooltip=c("text"))
```

## Open this in a new tab.
```{r pairs, fig.width=30, fig.height=30}
df_casted <- dcast(dfm, url ~ measurement, value.var="seconds")
df_casted$url = NULL # Don't care about URL
df_casted <- na.omit(df_casted)

generate_scatter <- function(data, mapping, ...){
  ggplot(data = data, mapping = mapping) + 
    geom_point(alpha = 0.1) +
    scale_x_continuous(trans = "log10", breaks=b, labels=b) +
    scale_y_continuous(trans = "log10", breaks=b, labels=b)
}

generate_density <- function(data, mapping, ...) {
    r = ggplot(data = data, mapping = mapping) + 
      geom_density(..., fill="red", color=NA) +
      scale_x_continuous(trans = "log10", breaks=b, labels=b)
    r
}

p = ggpairs(df_casted, 
      lower=list(continuous = generate_scatter),
      upper = list(continuous = wrap(ggally_cor, size = 10)),       
      diag=list(continuous = generate_density))

print(p)
```