just playing with julia data

playing.jl

Pkg.add("Gadfly")
using Gadfly

# Random cool meta stuff
print(code_typed(print, ()))
:(2 + 2) # quote expressions to see AST
print(code_lowered(print, ())) # see the lowered AST
# returns an array because some functions return multiple values

Pkg.add("Cairo")
Pkg.add("DataFrames")
Pkg.add("RDatasets")

using DataFrames
using RDatasets

iris = dataset("datasets", "iris")

# DataFrames deal with missing data smartly
v = [0.5, 0.6, 0.7, 0.9]
mean(v)

# NA Type like Null, poisons other values, like NaN for floating point numbers
 
1 + NA
1 > NA
typeof(NA)
isna(NA)

NA == NA
NaN == NaN
NA == 1

# v = [0.5, 0.6, 0.7, NA, 0.9] # can't do this. use DataArray to store Ts and NAs side by side
# dv = DataArray([1,2,3])
# you can add NAs *after* you create the DataArray, not init with
dv = @data([NA,1,2,3])
dv[1] = NA
join(dv, "::")
mean(dv)
mean(dropna(dv))

mean(array(dv, 1)) # subsitution all NAs with 1
2dv

# convert DataArray into Array
convert(Array, dropna(dv))

# a bunch of convenience constuctors coming from matlab heritage
DataArray(zeros(Float64, 8))
ones(Float64, 10)
falses(10)
trues(6)
eye(3) # identity matrix
diagm(4) # diagonal matrix

# use this @data macro to make DataArrays
foo = @data(["John Smith", "Jane Doe"])
@data(ones(10))

# MATRIX MATH
a = DataArray(eye(10))
a[5,5] = NA
a = a*a
print(a)

print(a.na) # this is a seperate matrix keeping track of NAs

# singular value decomposition
svd(a)

# fast fourier transform...
# fft(zeros(10))

# Dealing with Heterogenous Data
# (each type of column might be different)
# it's like a very simple implementation of a relation database
a = [1:5]
a[1:3]
a[1:end]
# a[0:3] 1 indexed!
in(10, a)
in(2, a)


# A and B are our columns
df = DataFrame(A = 1:4, B=["M", "F", "F", "M"])
df[:A]
size(df[:B])
size(df)
(4,2)[1]
nrows = size(df, 1)
ncols = size(df, 2)

head(df,2) # look at the first two rows
tail(df, 2) # look at the last two rows

# get columns
df[1] # same as DF[:A]
df[:A] == df[1]
mean(df[:A])

# get rows
df[2, :]

describe(df)
# Type    ASCIIString
# NAs     0
# NA%     0.0%
# Unique  2

cumsum(df[:A])
# do to each column
df = DataFrame(A = 1:4, B = 5:8)
colwise(cumsum, df)

# read csv etc http://dataframesjl.readthedocs.org/en/latest/io.html
# df = readtable("data.csv")
# write to csv
# writetable("output.csv", df)