# First
html <- readLines('http://www.forbes.com/wealth/celebrities/list')

# Alternatively
html <- readLines('http://www.stat.cmu.edu/~cshalizi/statcomp/lectures/24/celebrity100.html')

# Start with one case: Tiger Woods
# How is his data encoded in the html?
i <- grep('Tiger Woods', html)
cat(html[(i-4):(i+11)])

# Another case: Lady Gaga
i <- grep('Lady Gaga', html)[2]
cat(html[(i-4):(i+11)])

# Make one long string
html <- paste(html, collapse = '\n')

# ==============================================================
# = After much debugging, we arrive at this regular expression =
# ==============================================================
pat <- paste(
    '<tr>',
    '<td class="rank">(\\d+)</td>',
    '<td class="company"><a rel="[^"]*"></a>',
    '<img src="[^"]*" alt="[^"]*" />',
    '<h3>([^>]+)</h3></td>',
    '<td>\\$([[:digit:]]+)\\s*M</td>',
    '<td class="smallrank">(\\d+)</td>',
    '<td class="smallrank">(\\d+)</td>',
    '<td class="smallrank">(\\d+)</td>',
    '<td class="smallrank">(\\d+)</td>',
    '<td class="smallrank">(\\d+)</td>',
    '</tr>',
sep = '\\s*')

# First pass: extract disjoint cases 
m <- gregexpr(pat, html, ignore.case = TRUE)
x <- regmatches(html, m)
x <- do.call(c, x)

# Second pass: extract capture groups
m <- regexec(pat, x, ignore.case = TRUE)
celebs <- regmatches(x, m)

# Put it all together
library(plyr)
df <- ldply(celebs, function(x) data.frame(
        rank = as.numeric(x[2]), 
        name = gsub('^\\s*|\\s*$', '', x[3]),
        pay = as.numeric(x[4]),
        money.rank = as.numeric(x[5]),
        tvradio.rank = as.numeric(x[6]),
        press.rank = as.numeric(x[7]),
        web.rank = as.numeric(x[8]),
        social.rank = as.numeric(x[9]),
stringsAsFactors = FALSE))

# ======================
# = Some Data Analysis =
# ======================
library(ggplot2)

qplot(x = web.rank, y = pay, data = df)
qplot(x = web.rank, y = pay, data = df, log = 'y')