Class: Rover::DataFrame

Inherits:
Object
  • Object
show all
Defined in:
lib/rover/data_frame.rb

Instance Method Summary collapse

Constructor Details

#initialize(*args) ⇒ DataFrame

Returns a new instance of DataFrame.



3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
 
# File 'lib/rover/data_frame.rb', line 3

def initialize(*args)
  data, options = process_args(args)

  @vectors = {}
  types = (options[:types] || {}).transform_keys(&:to_s)

  if data.is_a?(DataFrame)
    data.vectors.each do |k, v|
      @vectors[k] = v
    end
  elsif data.is_a?(Hash)
    data.to_h.each do |k, v|
      @vectors[k] =
        if v.respond_to?(:to_a)
          Vector.new(v, type: types[k.to_s])
        else
          v
        end
    end

    # handle scalars
    size = @vectors.values.find { |v| v.is_a?(Vector) }&.size || 1
    @vectors.each_key do |k|
      @vectors[k] = to_vector(@vectors[k], size: size, type: types[k.to_s])
    end
  elsif data.is_a?(Array)
    vectors = {}
    raise ArgumentError, "Array elements must be hashes" unless data.all?(Hash)
    keys = data.flat_map(&:keys).uniq
    keys.each do |k|
      vectors[k] = []
    end
    data.each do |d|
      keys.each do |k|
        vectors[k] << d[k]
      end
    end
    vectors.each do |k, v|
      @vectors[k] = to_vector(v, type: types[k.to_s])
    end
  elsif defined?(ActiveRecord) && (data.is_a?(ActiveRecord::Relation) || (data.is_a?(Class) && data < ActiveRecord::Base) || data.is_a?(ActiveRecord::Result))
    result = data.is_a?(ActiveRecord::Result) ? data : data.connection_pool.with_connection { |c| c.select_all(data.all.to_sql) }
    result.columns.each_with_index do |k, i|
      @vectors[k] = to_vector(result.rows.map { |r| r[i] }, type: types[k.to_s])
    end
  else
    raise ArgumentError, "Cannot cast to data frame: #{data.class.name}"
  end

  # check keys
  @vectors.each_key do |k|
    check_key(k)
  end

  # TODO check for duplicate keys
  @vectors.transform_keys!(&:to_s)

  # check sizes
  sizes = @vectors.values.map(&:size).uniq
  if sizes.size > 1
    raise ArgumentError, "Different sizes: #{sizes}"
  end
end

Instance Method Details

#+(other) ⇒ Object 



385
386
387
 
# File 'lib/rover/data_frame.rb', line 385

def +(other)
  dup.concat(other)
end

#==(other) ⇒ Object

don't check types



426
427
428
429
430
 
# File 'lib/rover/data_frame.rb', line 426

def ==(other)
  size == other.size &&
  keys == other.keys &&
  keys.all? { |k| self[k].to_numo == other[k].to_numo }
end

#[](where) ⇒ Object 



67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
 
# File 'lib/rover/data_frame.rb', line 67

def [](where)
  # TODO use path for all vectors in 2.0
  if (where.is_a?(Vector) && (where.type == :bool || Vector::INTEGER_TYPES.include?(where.type))) || where.is_a?(Numeric) || where.is_a?(Range) || (where.is_a?(Array) && where.all?(Integer))
    new_vectors = {}
    @vectors.each do |k, v|
      new_vectors[k] = v[where]
    end
    DataFrame.new(new_vectors)
  elsif where.is_a?(Array)
    # TODO raise error in 2.0
    unless where.all? { |v| v.is_a?(String) || v.is_a?(Symbol) }
      warn "[rover] Unsupported selector"
    end

    # multiple columns
    df = DataFrame.new
    where.each do |k|
      check_column(k)
      df[k] = @vectors[k.to_s]
    end
    df
  else
    # TODO raise error in 2.0
    unless where.is_a?(String) || where.is_a?(Symbol)
      warn "[rover] Unsupported selector"
    end

    # single column
    @vectors[where.to_s]
  end
end

#[]=(k, v) ⇒ Object

Raises:

  • (ArgumentError) 


116
117
118
119
120
121
 
# File 'lib/rover/data_frame.rb', line 116

def []=(k, v)
  check_key(k)
  v = to_vector(v, size: size)
  raise ArgumentError, "Size mismatch (given #{v.size}, expected #{size})" if @vectors.any? && v.size != size
  @vectors[k.to_s] = v
end

#any? ⇒ Boolean

should this check for columns as well?

Returns:

  • (Boolean) 


130
131
132
 
# File 'lib/rover/data_frame.rb', line 130

def any?
  size > 0
end

#clear ⇒ Object 



139
140
141
 
# File 'lib/rover/data_frame.rb', line 139

def clear
  @vectors.clear
end

#concat(other) ⇒ Object

in-place, like Array#concat TODO make more performant

Raises:

  • (ArgumentError) 


391
392
393
394
395
396
397
398
399
400
401
402
 
# File 'lib/rover/data_frame.rb', line 391

def concat(other)
  raise ArgumentError, "Must be a data frame" unless other.is_a?(DataFrame)

  size = self.size
  vectors.each do |k, v|
    @vectors[k] = Vector.new(v.to_a + (other[k] ? other[k].to_a : Array.new(other.size)))
  end
  (other.vector_names - vector_names).each do |k|
    @vectors[k] = Vector.new(Array.new(size) + other[k].to_a)
  end
  self
end

#deep_dup ⇒ Object 



377
378
379
380
381
382
383
 
# File 'lib/rover/data_frame.rb', line 377

def deep_dup
  df = DataFrame.new
  @vectors.each do |k, v|
    df[k] = v.dup
  end
  df
end

#delete(key) ⇒ Object 



165
166
167
 
# File 'lib/rover/data_frame.rb', line 165

def delete(key)
  @vectors.delete(key.to_s)
end

#each_row ⇒ Object 



99
100
101
102
103
104
105
 
# File 'lib/rover/data_frame.rb', line 99

def each_row
  return enum_for(:each_row) unless block_given?

  size.times do |i|
    yield @vectors.map { |k, v| [k, v[i]] }.to_h
  end
end

#empty? ⇒ Boolean

should this check for columns as well?

Returns:

  • (Boolean) 


135
136
137
 
# File 'lib/rover/data_frame.rb', line 135

def empty?
  size == 0
end

#except(*keys) ⇒ Object 



169
170
171
 
# File 'lib/rover/data_frame.rb', line 169

def except(*keys)
  dup.except!(*keys)
end

#except!(*keys) ⇒ Object 



173
174
175
176
177
178
 
# File 'lib/rover/data_frame.rb', line 173

def except!(*keys)
  keys.each do |key|
    delete(key)
  end
  self
end

#first(n = 1) ⇒ Object 



192
193
194
195
196
197
198
 
# File 'lib/rover/data_frame.rb', line 192

def first(n = 1)
  new_vectors = {}
  @vectors.each do |k, v|
    new_vectors[k] = v.first(n)
  end
  DataFrame.new(new_vectors)
end

#group(*columns) ⇒ Object 



366
367
368
 
# File 'lib/rover/data_frame.rb', line 366

def group(*columns)
  Group.new(self, columns.flatten.map(&:to_s))
end

#head(n = 5) ⇒ Object 



184
185
186
 
# File 'lib/rover/data_frame.rb', line 184

def head(n = 5)
  first(n)
end

#include?(key) ⇒ Boolean

Returns:

  • (Boolean) 


180
181
182
 
# File 'lib/rover/data_frame.rb', line 180

def include?(key)
  @vectors.include?(key.to_s)
end

#inner_join(other, on: nil) ⇒ Object

see join for options



416
417
418
 
# File 'lib/rover/data_frame.rb', line 416

def inner_join(other, on: nil)
  join(other, on: on, how: "inner")
end

#inspect ⇒ Object Also known as: to_s

TODO handle long text better



316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
 
# File 'lib/rover/data_frame.rb', line 316

def inspect
  return "#<Rover::DataFrame>" if keys.empty?

  lines = []
  line_start = 0
  spaces = 2

  summarize = size >= 30

  @vectors.each do |k, v|
    v = summarize ? v.first(5).to_a + ["..."] + v.last(5).to_a : v.to_a
    width = ([k] + v).map { |v2| v2.to_s.size }.max
    width = 3 if width < 3

    if lines.empty? || lines[-2].sum { |l| l.size + spaces } + width > 120
      line_start = lines.size
      lines << []
      v.size.times do |i|
        lines << []
      end
      lines << []
    end

    lines[line_start] << "%#{width}s" % k.to_s
    v.each_with_index do |v2, i|
      lines[line_start + 1 + i] << "%#{width}s" % v2.to_s
    end
  end

  lines.pop
  lines.map { |l| l.join(" " * spaces) }.join("\n")
end

#keys ⇒ Object Also known as: names, vector_names 



147
148
149
 
# File 'lib/rover/data_frame.rb', line 147

def keys
  @vectors.keys
end

#last(n = 1) ⇒ Object 



200
201
202
203
204
205
206
 
# File 'lib/rover/data_frame.rb', line 200

def last(n = 1)
  new_vectors = {}
  @vectors.each do |k, v|
    new_vectors[k] = v.last(n)
  end
  DataFrame.new(new_vectors)
end

#left_join(other, on: nil) ⇒ Object

see join for options



421
422
423
 
# File 'lib/rover/data_frame.rb', line 421

def left_join(other, on: nil)
  join(other, on: on, how: "left")
end

#merge(other) ⇒ Object 



404
405
406
 
# File 'lib/rover/data_frame.rb', line 404

def merge(other)
  dup.merge!(other)
end

#merge!(other) ⇒ Object 



408
409
410
411
412
413
 
# File 'lib/rover/data_frame.rb', line 408

def merge!(other)
  other.vectors.each do |k, v|
    self[k] = v
  end
  self
end

#one_hot(drop: false) ⇒ Object

TODO raise error when collision



235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
 
# File 'lib/rover/data_frame.rb', line 235

def one_hot(drop: false)
  df = DataFrame.new
  vectors.each do |k, v|
    if v.to_numo.is_a?(Numo::RObject)
      df.merge!(v.one_hot(drop: drop, prefix: "#{k}_"))
    else
      df[k] = v
    end
  end
  df
rescue ArgumentError => e
  if e.message == "All elements must be strings"
    # better error message
    raise ArgumentError, "All elements must be numeric or strings"
  end
  raise e
end

#plot(x = nil, y = nil, type: nil, group: nil, stacked: nil) ⇒ Object

Raises:

  • (ArgumentError) 


432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
 
# File 'lib/rover/data_frame.rb', line 432

def plot(x = nil, y = nil, type: nil, group: nil, stacked: nil)
  require "vega"

  raise ArgumentError, "Must specify columns" if keys.size != 2 && (!x || !y)
  x ||= keys[0]
  y ||= keys[1]
  type ||= begin
    if self[x].numeric? && self[y].numeric?
      "scatter"
    elsif types[x] == :object && self[y].numeric?
      "column"
    else
      raise "Cannot determine type. Use the type option."
    end
  end
  data = self[group.nil? ? [x, y] : [x, y, group]]

  case type
  when "line", "area"
    x_type =
      if data[x].numeric?
        "quantitative"
      elsif data[x].all? { |v| v.is_a?(Date) || v.is_a?(Time) }
        "temporal"
      else
        "nominal"
      end

    scale = x_type == "temporal" ? {type: "utc"} : {}
    encoding = {
      x: {field: x, type: x_type, scale: scale},
      y: {field: y, type: "quantitative"}
    }
    encoding[:color] = {field: group} if group

    Vega.lite
      .data(data)
      .mark(type: type, tooltip: true, interpolate: "cardinal", point: {size: 60})
      .encoding(encoding)
      .config(axis: {labelFontSize: 12})
  when "pie"
    raise ArgumentError, "Cannot use group option with pie chart" unless group.nil?

    Vega.lite
      .data(data)
      .mark(type: "arc", tooltip: true)
      .encoding(
        color: {field: x, type: "nominal", sort: "none", axis: {title: nil}, legend: {labelFontSize: 12}},
        theta: {field: y, type: "quantitative"}
      )
      .view(stroke: nil)
  when "column"
    encoding = {
      x: {field: x, type: "nominal", sort: "none", axis: {labelAngle: 0}},
      y: {field: y, type: "quantitative"}
    }
    if group
      encoding[:color] = {field: group}
      encoding[:xOffset] = {field: group} unless stacked
    end

    Vega.lite
      .data(data)
      .mark(type: "bar", tooltip: true)
      .encoding(encoding)
      .config(axis: {labelFontSize: 12})
  when "bar"
    encoding = {
      # TODO determine label angle
      y: {field: x, type: "nominal", sort: "none", axis: {labelAngle: 0}},
      x: {field: y, type: "quantitative"}
    }
    if group
      encoding[:color] = {field: group}
      encoding[:yOffset] = {field: group} unless stacked
    end

    Vega.lite
      .data(data)
      .mark(type: "bar", tooltip: true)
      .encoding(encoding)
      .config(axis: {labelFontSize: 12})
  when "scatter"
    encoding = {
      x: {field: x, type: "quantitative", scale: {zero: false}},
      y: {field: y, type: "quantitative", scale: {zero: false}},
      size: {value: 60}
    }
    encoding[:color] = {field: group} if group

    Vega.lite
      .data(data)
      .mark(type: "circle", tooltip: true)
      .encoding(encoding)
      .config(axis: {labelFontSize: 12})
  else
    raise ArgumentError, "Invalid type: #{type}"
  end
end

#rename(mapping) ⇒ Object 



153
154
155
156
157
158
159
160
161
162
163
 
# File 'lib/rover/data_frame.rb', line 153

def rename(mapping)
  mapping.each_key do |k|
    check_column(k)
  end
  mapping = mapping.to_h { |k, v| [k.to_s, v.to_s] }
  # use transform_keys! to preserve order
  @vectors.transform_keys! do |k|
    mapping[k] || k
  end
  self
end

#sample(*args, **kwargs) ⇒ Object 



208
209
210
211
212
 
# File 'lib/rover/data_frame.rb', line 208

def sample(*args, **kwargs)
  # TODO make more efficient
  indexes = (0...size).to_a.sample(*args, **kwargs)
  self[indexes]
end

#shape ⇒ Object 



143
144
145
 
# File 'lib/rover/data_frame.rb', line 143

def shape
  [size, @vectors.size]
end

#size ⇒ Object Also known as: length, count 



123
124
125
 
# File 'lib/rover/data_frame.rb', line 123

def size
  @vectors.values.first&.size || 0
end

#sort_by(&block) ⇒ Object 



362
363
364
 
# File 'lib/rover/data_frame.rb', line 362

def sort_by(&block)
  dup.sort_by!(&block)
end

#sort_by! ⇒ Object 



350
351
352
353
354
355
356
357
358
359
360
 
# File 'lib/rover/data_frame.rb', line 350

def sort_by!
  indexes =
    size.times.sort_by do |i|
      yield @vectors.map { |k, v| [k, v[i]] }.to_h
    end

  @vectors.each do |k, v|
    self[k] = v.to_numo.at(indexes)
  end
  self
end

#tail(n = 5) ⇒ Object 



188
189
190
 
# File 'lib/rover/data_frame.rb', line 188

def tail(n = 5)
  last(n)
end

#to_a ⇒ Object 



214
215
216
217
218
219
220
 
# File 'lib/rover/data_frame.rb', line 214

def to_a
  a = []
  each_row do |row|
    a << row
  end
  a
end

#to_csv ⇒ Object 



253
254
255
256
257
258
259
260
261
262
263
 
# File 'lib/rover/data_frame.rb', line 253

def to_csv
  require "csv"

  CSV.generate do |csv|
    csv << keys
    numo = vectors.values.map(&:to_numo)
    size.times do |i|
      csv << numo.map { |n| n[i] }
    end
  end
end

#to_h ⇒ Object 



222
223
224
225
226
227
228
 
# File 'lib/rover/data_frame.rb', line 222

def to_h
  hsh = {}
  @vectors.each do |k, v|
    hsh[k] = v.to_a
  end
  hsh
end

#to_html ⇒ Object

for IRuby



304
305
306
307
308
309
310
311
312
313
 
# File 'lib/rover/data_frame.rb', line 304

def to_html
  require "iruby"

  if size > 7
    # pass 8 rows so maxrows is applied
    IRuby::HTML.table((self[0..4] + self[-4..-1]).to_h, maxrows: 7)
  else
    IRuby::HTML.table(to_h)
  end
end

#to_numo ⇒ Object 



230
231
232
 
# File 'lib/rover/data_frame.rb', line 230

def to_numo
  Numo::NArray.column_stack(vectors.values.map(&:to_numo))
end

#to_parquet ⇒ Object 



265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
 
# File 'lib/rover/data_frame.rb', line 265

def to_parquet
  require "parquet"

  schema = {}
  types.each do |name, type|
    schema[name] =
      case type
      when :int64
        :int64
      when :uint64
        :uint64
      when :float64
        :double
      when :float32
        :float
      when :bool
        :boolean
      when :object
        if @vectors[name].all?(String)
          :string
        else
          raise "Unknown type"
        end
      else
        type
      end
  end
  # TODO improve performance
  raw_records = []
  size.times do |i|
    raw_records << @vectors.map { |_, v| v[i] }
  end
  table = Arrow::Table.new(schema, raw_records)
  buffer = Arrow::ResizableBuffer.new(1024)
  table.save(buffer, format: :parquet)
  buffer.data.to_s
end

#types ⇒ Object 



112
113
114
 
# File 'lib/rover/data_frame.rb', line 112

def types
  @vectors.map { |k, v| [k, v.type] }.to_h
end

#vectors ⇒ Object

dup to prevent direct modification of keys



108
109
110
 
# File 'lib/rover/data_frame.rb', line 108

def vectors
  @vectors.dup
end