ul-wukong 4.1.0
Sign up to get free protection for your applications and to get access to all the features.
- checksums.yaml +15 -0
- data/.gitignore +60 -0
- data/.gitmodules +6 -0
- data/.rspec +2 -0
- data/.travis.yml +19 -0
- data/.yardopts +6 -0
- data/CHANGELOG.md +7 -0
- data/Gemfile +17 -0
- data/Guardfile +12 -0
- data/LICENSE.md +95 -0
- data/NOTES-travis.md +31 -0
- data/README-old.md +422 -0
- data/README.md +1308 -0
- data/Rakefile +28 -0
- data/TODO.md +99 -0
- data/bin/cutc +30 -0
- data/bin/cuttab +5 -0
- data/bin/greptrue +6 -0
- data/bin/md5sort +20 -0
- data/bin/setcat +11 -0
- data/bin/tabchar +5 -0
- data/bin/uniq-ord +59 -0
- data/bin/uniqc +3 -0
- data/bin/wu +34 -0
- data/bin/wu-clean-encoding +31 -0
- data/bin/wu-date +13 -0
- data/bin/wu-datetime +13 -0
- data/bin/wu-hist +3 -0
- data/bin/wu-lign +186 -0
- data/bin/wu-local +4 -0
- data/bin/wu-plus +9 -0
- data/bin/wu-source +5 -0
- data/bin/wu-sum +31 -0
- data/diagrams/wu_local.dot +39 -0
- data/diagrams/wu_local.dot.png +0 -0
- data/examples/Gemfile +38 -0
- data/examples/README.md +9 -0
- data/examples/basic/string_reverser.rb +23 -0
- data/examples/basic/tiny_count.rb +8 -0
- data/examples/basic/word_count/accumulator.rb +26 -0
- data/examples/basic/word_count/tokenizer.rb +13 -0
- data/examples/basic/word_count/word_count.rb +6 -0
- data/examples/dataflow/scraper_macro_flow.rb +28 -0
- data/examples/deploy_pack/Gemfile +6 -0
- data/examples/deploy_pack/README.md +6 -0
- data/examples/deploy_pack/a/b/c/.gitkeep +0 -0
- data/examples/deploy_pack/app/processors/string_reverser.rb +5 -0
- data/examples/deploy_pack/config/environment.rb +1 -0
- data/examples/dsl/dataflow/fibonacci_series.rb +101 -0
- data/examples/dsl/dataflow/scraper_macro_flow.rb +28 -0
- data/examples/dsl/dataflow/simple.rb +12 -0
- data/examples/dsl/dataflow/telegram.rb +45 -0
- data/examples/dsl/workflow/cherry_pie.dot +97 -0
- data/examples/dsl/workflow/cherry_pie.md +104 -0
- data/examples/dsl/workflow/cherry_pie.png +0 -0
- data/examples/dsl/workflow/cherry_pie.rb +101 -0
- data/examples/empty/.gitkeep +0 -0
- data/examples/examples_helper.rb +9 -0
- data/examples/geo.rb +4 -0
- data/examples/geo/geo_grids.numbers +0 -0
- data/examples/geo/geolocated.rb +331 -0
- data/examples/geo/quadtile.rb +69 -0
- data/examples/geo/spec/geolocated_spec.rb +247 -0
- data/examples/geo/tile_fetcher.rb +77 -0
- data/examples/graph/implied_geolocation/README.md +63 -0
- data/examples/graph/minimum_spanning_tree/airfares_graphviz.rb +73 -0
- data/examples/improver/tweet_summary.rb +73 -0
- data/examples/loadable.rb +2 -0
- data/examples/munging/airline_flights/airline_flights.rake +83 -0
- data/examples/munging/airline_flights/airplane.rb +0 -0
- data/examples/munging/airline_flights/airport_id_unification.rb +129 -0
- data/examples/munging/airline_flights/airport_ok_chars.rb +4 -0
- data/examples/munging/airline_flights/indexable.rb +75 -0
- data/examples/munging/airline_flights/indexable_spec.rb +90 -0
- data/examples/munging/airline_flights/reconcile_airports.rb +142 -0
- data/examples/munging/airline_flights/tasks.rake +83 -0
- data/examples/munging/airline_flights/topcities.rb +167 -0
- data/examples/munging/geo/geo_json.rb +54 -0
- data/examples/munging/geo/geo_models.rb +69 -0
- data/examples/munging/geo/geonames_models.rb +107 -0
- data/examples/munging/geo/iso_codes.rb +172 -0
- data/examples/munging/geo/reconcile_countries.rb +124 -0
- data/examples/munging/geo/tasks.rake +71 -0
- data/examples/munging/wikipedia/articles/extract_articles-parsed.rb +79 -0
- data/examples/munging/wikipedia/articles/extract_articles-templated.rb +136 -0
- data/examples/munging/wikipedia/articles/textualize_articles.rb +54 -0
- data/examples/munging/wikipedia/articles/verify_structure.rb +43 -0
- data/examples/munging/wikipedia/articles/wp2txt-LICENSE.txt +22 -0
- data/examples/munging/wikipedia/articles/wp2txt_article.rb +259 -0
- data/examples/munging/wikipedia/articles/wp2txt_utils.rb +452 -0
- data/examples/munging/wikipedia/dbpedia/dbpedia_common.rb +5 -0
- data/examples/munging/wikipedia/dbpedia/dbpedia_extract_geocoordinates.rb +78 -0
- data/examples/munging/wikipedia/dbpedia/extract_links-cruft.rb +66 -0
- data/examples/munging/wikipedia/dbpedia/extract_links.rb +260 -0
- data/examples/munging/wikipedia/dbpedia/sameas_extractor.rb +20 -0
- data/examples/rake_helper.rb +97 -0
- data/examples/ruby_project/Gemfile +6 -0
- data/examples/ruby_project/README.md +6 -0
- data/examples/ruby_project/a/b/c/.gitkeep +0 -0
- data/examples/server_logs/geo_ip_mapping/munge_geolite.rb +82 -0
- data/examples/server_logs/logline.rb +95 -0
- data/examples/server_logs/models.rb +66 -0
- data/examples/server_logs/page_counts.pig +48 -0
- data/examples/server_logs/server_logs-01-parse-script.rb +13 -0
- data/examples/server_logs/server_logs-02-histograms-full.rb +33 -0
- data/examples/server_logs/server_logs-02-histograms-mapper.rb +14 -0
- data/examples/server_logs/server_logs-03-breadcrumbs-full.rb +71 -0
- data/examples/server_logs/server_logs-04-page_page_edges-full.rb +40 -0
- data/examples/serverlogs/geo_ip_mapping/munge_geolite.rb +82 -0
- data/examples/serverlogs/models/logline.rb +102 -0
- data/examples/serverlogs/parser/apache_parser_widget.rb +46 -0
- data/examples/serverlogs/visit_paths/common.rb +4 -0
- data/examples/serverlogs/visit_paths/page_counts.pig +48 -0
- data/examples/serverlogs/visit_paths/serverlogs-01-parse-script.rb +11 -0
- data/examples/serverlogs/visit_paths/serverlogs-02-histograms-full.rb +31 -0
- data/examples/serverlogs/visit_paths/serverlogs-02-histograms-mapper.rb +12 -0
- data/examples/serverlogs/visit_paths/serverlogs-03-breadcrumbs-full.rb +67 -0
- data/examples/serverlogs/visit_paths/serverlogs-04-page_page_edges-full.rb +38 -0
- data/examples/splitter.rb +94 -0
- data/examples/string_reverser.rb +7 -0
- data/examples/text/pig_latin/pig_latinizer.rb +35 -0
- data/examples/text/pig_latin/pig_latinizer_widget.rb +16 -0
- data/examples/text/regional_flavor/README.md +14 -0
- data/examples/text/regional_flavor/article_wordbags.pig +39 -0
- data/examples/text/regional_flavor/j01-article_wordbags.rb +4 -0
- data/examples/text/regional_flavor/simple_pig_script.pig +27 -0
- data/examples/twitter.rb +5 -0
- data/lib/hanuman.rb +36 -0
- data/lib/hanuman/graph.rb +97 -0
- data/lib/hanuman/graphvizzer.rb +206 -0
- data/lib/hanuman/graphvizzer/gv_models.rb +161 -0
- data/lib/hanuman/graphvizzer/gv_presenter.rb +97 -0
- data/lib/hanuman/link.rb +35 -0
- data/lib/hanuman/registry.rb +46 -0
- data/lib/hanuman/stage.rb +128 -0
- data/lib/hanuman/tree.rb +67 -0
- data/lib/wu/geo.rb +4 -0
- data/lib/wu/geo/geo_grids.numbers +0 -0
- data/lib/wu/geo/geolocated.rb +331 -0
- data/lib/wu/geo/quadtile.rb +69 -0
- data/lib/wu/graph/union_find.rb +62 -0
- data/lib/wu/model/reconcilable.rb +63 -0
- data/lib/wu/munging.rb +71 -0
- data/lib/wu/social/models/twitter.rb +31 -0
- data/lib/wu/wikipedia/models.rb +20 -0
- data/lib/wukong.rb +54 -0
- data/lib/wukong/dataflow.rb +43 -0
- data/lib/wukong/doc_helpers.rb +14 -0
- data/lib/wukong/doc_helpers/dataflow_handler.rb +29 -0
- data/lib/wukong/doc_helpers/field_handler.rb +91 -0
- data/lib/wukong/doc_helpers/processor_handler.rb +29 -0
- data/lib/wukong/driver.rb +214 -0
- data/lib/wukong/driver/event_machine_driver.rb +15 -0
- data/lib/wukong/driver/wiring.rb +68 -0
- data/lib/wukong/local.rb +42 -0
- data/lib/wukong/local/runner.rb +96 -0
- data/lib/wukong/local/stdio_driver.rb +104 -0
- data/lib/wukong/logger.rb +102 -0
- data/lib/wukong/model/faker.rb +136 -0
- data/lib/wukong/model/flatpack_parser/flat.rb +60 -0
- data/lib/wukong/model/flatpack_parser/flatpack.rb +4 -0
- data/lib/wukong/model/flatpack_parser/lang.rb +46 -0
- data/lib/wukong/model/flatpack_parser/parser.rb +55 -0
- data/lib/wukong/model/flatpack_parser/tokens.rb +130 -0
- data/lib/wukong/plugin.rb +48 -0
- data/lib/wukong/processor.rb +110 -0
- data/lib/wukong/rake_helper.rb +6 -0
- data/lib/wukong/runner.rb +169 -0
- data/lib/wukong/runner/boot_sequence.rb +123 -0
- data/lib/wukong/runner/code_loader.rb +52 -0
- data/lib/wukong/runner/command_runner.rb +44 -0
- data/lib/wukong/runner/deploy_pack_loader.rb +75 -0
- data/lib/wukong/runner/help_message.rb +42 -0
- data/lib/wukong/source.rb +33 -0
- data/lib/wukong/source/source_driver.rb +74 -0
- data/lib/wukong/source/source_runner.rb +38 -0
- data/lib/wukong/spec_helpers.rb +74 -0
- data/lib/wukong/spec_helpers/integration_tests.rb +150 -0
- data/lib/wukong/spec_helpers/integration_tests/integration_test_matchers.rb +207 -0
- data/lib/wukong/spec_helpers/integration_tests/integration_test_runner.rb +97 -0
- data/lib/wukong/spec_helpers/shared_examples.rb +22 -0
- data/lib/wukong/spec_helpers/unit_tests.rb +135 -0
- data/lib/wukong/spec_helpers/unit_tests/unit_test_driver.rb +132 -0
- data/lib/wukong/spec_helpers/unit_tests/unit_test_matchers.rb +169 -0
- data/lib/wukong/spec_helpers/unit_tests/unit_test_runner.rb +60 -0
- data/lib/wukong/version.rb +3 -0
- data/lib/wukong/widget/echo.rb +55 -0
- data/lib/wukong/widget/extract.rb +122 -0
- data/lib/wukong/widget/filters.rb +452 -0
- data/lib/wukong/widget/logger.rb +56 -0
- data/lib/wukong/widget/operators.rb +82 -0
- data/lib/wukong/widget/reducers.rb +10 -0
- data/lib/wukong/widget/reducers/accumulator.rb +73 -0
- data/lib/wukong/widget/reducers/bin.rb +368 -0
- data/lib/wukong/widget/reducers/count.rb +73 -0
- data/lib/wukong/widget/reducers/group.rb +128 -0
- data/lib/wukong/widget/reducers/group_concat.rb +98 -0
- data/lib/wukong/widget/reducers/improver.rb +71 -0
- data/lib/wukong/widget/reducers/join_xml.rb +37 -0
- data/lib/wukong/widget/reducers/moments.rb +72 -0
- data/lib/wukong/widget/reducers/sort.rb +180 -0
- data/lib/wukong/widget/reducers/uniq.rb +91 -0
- data/lib/wukong/widget/serializers.rb +317 -0
- data/lib/wukong/widget/utils.rb +46 -0
- data/lib/wukong/widgets.rb +7 -0
- data/spec/examples/dataflow/fibonacci_series_spec.rb +18 -0
- data/spec/examples/dataflow/parse_apache_logs_spec.rb +8 -0
- data/spec/examples/dataflow/parsing_spec.rb +14 -0
- data/spec/examples/dataflow/simple_spec.rb +34 -0
- data/spec/examples/dataflow/telegram_spec.rb +43 -0
- data/spec/examples/graph/minimum_spanning_tree_spec.rb +34 -0
- data/spec/examples/munging/airline_flights/identifiers_spec.rb +16 -0
- data/spec/examples/munging/airline_flights_spec.rb +202 -0
- data/spec/examples/text/pig_latin_spec.rb +18 -0
- data/spec/examples/workflow/cherry_pie_spec.rb +36 -0
- data/spec/hanuman/graph_spec.rb +119 -0
- data/spec/hanuman/hanuman_spec.rb +10 -0
- data/spec/hanuman/registry_spec.rb +123 -0
- data/spec/hanuman/stage_spec.rb +81 -0
- data/spec/hanuman/tree_spec.rb +119 -0
- data/spec/spec.opts +1 -0
- data/spec/spec_helper.rb +43 -0
- data/spec/support/example_test_helpers.rb +95 -0
- data/spec/support/hanuman_test_helpers.rb +92 -0
- data/spec/support/integration_helper.rb +38 -0
- data/spec/support/model_test_helpers.rb +115 -0
- data/spec/support/shared_context_for_graphs.rb +57 -0
- data/spec/support/shared_context_for_reducers.rb +37 -0
- data/spec/support/shared_examples_for_builders.rb +94 -0
- data/spec/support/shared_examples_for_shortcuts.rb +57 -0
- data/spec/wu/model/reconcilable_spec.rb +152 -0
- data/spec/wukong/dataflow_spec.rb +87 -0
- data/spec/wukong/driver_spec.rb +154 -0
- data/spec/wukong/local/runner_spec.rb +29 -0
- data/spec/wukong/local/stdio_driver_spec.rb +73 -0
- data/spec/wukong/local_spec.rb +6 -0
- data/spec/wukong/logger_spec.rb +49 -0
- data/spec/wukong/model/faker_spec.rb +132 -0
- data/spec/wukong/processor_spec.rb +21 -0
- data/spec/wukong/runner_spec.rb +132 -0
- data/spec/wukong/source_spec.rb +6 -0
- data/spec/wukong/widget/extract_spec.rb +101 -0
- data/spec/wukong/widget/filters_spec.rb +79 -0
- data/spec/wukong/widget/logger_spec.rb +23 -0
- data/spec/wukong/widget/operators_spec.rb +25 -0
- data/spec/wukong/widget/reducers/bin_spec.rb +92 -0
- data/spec/wukong/widget/reducers/count_spec.rb +11 -0
- data/spec/wukong/widget/reducers/group_spec.rb +21 -0
- data/spec/wukong/widget/reducers/join_xml_spec.rb +25 -0
- data/spec/wukong/widget/reducers/moments_spec.rb +36 -0
- data/spec/wukong/widget/reducers/sort_spec.rb +26 -0
- data/spec/wukong/widget/reducers/uniq_spec.rb +14 -0
- data/spec/wukong/widget/serializers_spec.rb +114 -0
- data/spec/wukong/widget/sink_spec.rb +19 -0
- data/spec/wukong/widget/source_spec.rb +65 -0
- data/spec/wukong/wu-local_spec.rb +109 -0
- data/spec/wukong/wu-source_spec.rb +32 -0
- data/spec/wukong/wu_spec.rb +14 -0
- data/spec/wukong/wukong_spec.rb +10 -0
- data/wukong.gemspec +35 -0
- metadata +465 -0
@@ -0,0 +1,73 @@
|
|
1
|
+
require_relative("accumulator")
|
2
|
+
|
3
|
+
module Wukong
|
4
|
+
class Processor
|
5
|
+
|
6
|
+
# A processor which counts the total number of its input records.
|
7
|
+
#
|
8
|
+
# On it's own, this widget is really just a poor man's `wc -l`.
|
9
|
+
# It's really intended to serve as a superclass for more complex
|
10
|
+
# accumulators.
|
11
|
+
#
|
12
|
+
# @example Count the total number of input records on the command-line.
|
13
|
+
#
|
14
|
+
# $ wc -l input
|
15
|
+
# 283 input
|
16
|
+
# $ cat input | wu-local count
|
17
|
+
# 283
|
18
|
+
class Count < Accumulator
|
19
|
+
|
20
|
+
description <<EOF
|
21
|
+
This processor counts the number of input records it receives.
|
22
|
+
|
23
|
+
$ wc -l input
|
24
|
+
283 input
|
25
|
+
$ cat input | wu-local count
|
26
|
+
283
|
27
|
+
|
28
|
+
This processor will not output any records until it receives its final
|
29
|
+
input record.
|
30
|
+
EOF
|
31
|
+
|
32
|
+
# The total size of the input recors.
|
33
|
+
attr_accessor :size
|
34
|
+
|
35
|
+
# Initializes the count to 0.
|
36
|
+
def setup
|
37
|
+
super()
|
38
|
+
@size = 0
|
39
|
+
end
|
40
|
+
|
41
|
+
# Accumulate a `record` by incrmenting the total size.
|
42
|
+
#
|
43
|
+
# @param [Object] record
|
44
|
+
def accumulate record
|
45
|
+
self.size += 1
|
46
|
+
end
|
47
|
+
|
48
|
+
# Keeps all records in the same group so that one count is
|
49
|
+
# emitted at the end.
|
50
|
+
#
|
51
|
+
# Overriding this method and returning different keys for
|
52
|
+
# different records is the beginning of constructing a "group
|
53
|
+
# by" type widget.
|
54
|
+
#
|
55
|
+
# @param [Object] record
|
56
|
+
# @return [:__first__group__]
|
57
|
+
# @see Group
|
58
|
+
def get_key record
|
59
|
+
:__first_group__
|
60
|
+
end
|
61
|
+
|
62
|
+
# Yields the total size.
|
63
|
+
#
|
64
|
+
# @yield [size]
|
65
|
+
# @yieldparam [Integer] size
|
66
|
+
def finalize
|
67
|
+
yield self.size
|
68
|
+
end
|
69
|
+
|
70
|
+
register
|
71
|
+
end
|
72
|
+
end
|
73
|
+
end
|
@@ -0,0 +1,128 @@
|
|
1
|
+
require_relative("../utils")
|
2
|
+
require_relative("count")
|
3
|
+
|
4
|
+
module Wukong
|
5
|
+
class Processor
|
6
|
+
|
7
|
+
# Groups sorted input records and emits each group with a count.
|
8
|
+
#
|
9
|
+
# Allows you to use several ways of extracting the key that
|
10
|
+
# defines the group.
|
11
|
+
#
|
12
|
+
# **Note:** The input records must be previously sorted by the
|
13
|
+
# same key used for grouping in order to ensure that groups are
|
14
|
+
# not split up.
|
15
|
+
#
|
16
|
+
# @example Group simple string values on the command-line.
|
17
|
+
#
|
18
|
+
# $ cat input
|
19
|
+
# apple
|
20
|
+
# cat
|
21
|
+
# banana
|
22
|
+
# apple
|
23
|
+
# ...
|
24
|
+
# $ cat input | wu-local sort | wu-local group --to=tsv
|
25
|
+
# apple 4
|
26
|
+
# banana 2
|
27
|
+
# cat 5
|
28
|
+
# ...
|
29
|
+
#
|
30
|
+
# @example Group using a nested key within a JSON string on the command-line
|
31
|
+
#
|
32
|
+
# $ cat input
|
33
|
+
# {"id": 1, "word": "apple" }
|
34
|
+
# {"id": 2, "word": "cat" }
|
35
|
+
# {"id": 3, "word": "banana"}
|
36
|
+
# ...
|
37
|
+
# $ cat input | wu-local sort --on=word | wu-local group --by=word --to=tsv
|
38
|
+
# apple 4
|
39
|
+
# banana 2
|
40
|
+
# cat 5
|
41
|
+
# ...
|
42
|
+
#
|
43
|
+
# A group fits nicely at the end of a dataflow. Since it requires
|
44
|
+
# a sort, it is blocking.
|
45
|
+
#
|
46
|
+
# @example Using a group at the end of a dataflow
|
47
|
+
#
|
48
|
+
# Wukong.dataflow(:makes_groups) do
|
49
|
+
# ... | sort(on: 'field') | group(by: 'field') | to_tsv
|
50
|
+
# end
|
51
|
+
#
|
52
|
+
# @see Sort
|
53
|
+
class Group < Count
|
54
|
+
|
55
|
+
description <<EOF
|
56
|
+
This processor groups consecutive input records that share the same
|
57
|
+
"group key". There are several ways to extract this group key from a
|
58
|
+
record.
|
59
|
+
|
60
|
+
NOTE: The input records must be previously sorted by the
|
61
|
+
same key used for grouping in order to ensure that groups are
|
62
|
+
not split up.
|
63
|
+
|
64
|
+
By default the input records themselves are used as their own group
|
65
|
+
keys, allowing to count identical values, a la `uniq -c`:
|
66
|
+
|
67
|
+
$ cat input
|
68
|
+
apple
|
69
|
+
cat
|
70
|
+
banana
|
71
|
+
apple
|
72
|
+
...
|
73
|
+
|
74
|
+
$ cat input | wu-local sort | wu-local group --to=tsv
|
75
|
+
apple 4
|
76
|
+
banana 2
|
77
|
+
cat 5
|
78
|
+
...
|
79
|
+
|
80
|
+
You can also group by some part of in input record:
|
81
|
+
|
82
|
+
$ cat input
|
83
|
+
{"id": 1, "word": "apple" }
|
84
|
+
{"id": 2, "word": "cat" }
|
85
|
+
{"id": 3, "word": "banana"}
|
86
|
+
...
|
87
|
+
|
88
|
+
$ cat input | wu-local sort --on==word | wu-local group --by=word --to=tsv
|
89
|
+
apple 4
|
90
|
+
banana 2
|
91
|
+
cat 5
|
92
|
+
...
|
93
|
+
|
94
|
+
This processor will not produce any output for a given group until it
|
95
|
+
sees the last record of that group.
|
96
|
+
EOF
|
97
|
+
|
98
|
+
include DynamicGet
|
99
|
+
field :by, Whatever, :doc => "Part of the record to group by"
|
100
|
+
|
101
|
+
# Get the key which defines the group for this `record`.
|
102
|
+
#
|
103
|
+
# @param [Object] record
|
104
|
+
# @return [Object]
|
105
|
+
def get_key(record)
|
106
|
+
get(self.by, record)
|
107
|
+
end
|
108
|
+
|
109
|
+
# Reset the size counter for new group.
|
110
|
+
#
|
111
|
+
# @param [Object] record
|
112
|
+
def start record
|
113
|
+
self.size = 0
|
114
|
+
end
|
115
|
+
|
116
|
+
# Yields the current group along with its size
|
117
|
+
#
|
118
|
+
# @yield [key, size]
|
119
|
+
# @yieldparam [Object] key the key defining the group
|
120
|
+
# @yieldparam [Integer] size the size of the group
|
121
|
+
def finalize
|
122
|
+
yield [key, size]
|
123
|
+
end
|
124
|
+
|
125
|
+
register
|
126
|
+
end
|
127
|
+
end
|
128
|
+
end
|
@@ -0,0 +1,98 @@
|
|
1
|
+
require_relative("group")
|
2
|
+
|
3
|
+
module Wukong
|
4
|
+
class Processor
|
5
|
+
|
6
|
+
# Concatenates the elements of a group, yielding the group key,
|
7
|
+
# the count, and its members.
|
8
|
+
#
|
9
|
+
# @example Concatenating elements of a group on the command-line.
|
10
|
+
#
|
11
|
+
# $ cat input
|
12
|
+
# {"id": 1, "parent_id": 4}
|
13
|
+
# {"id": 2, "parent_id": 3}
|
14
|
+
# {"id": 3, "parent_id": 3}
|
15
|
+
# ...
|
16
|
+
# $ cat input | wu-local group_concat --by=parent_id --to=tsv
|
17
|
+
# 4 1 {"id": 1, "parent_id": 4}
|
18
|
+
# 3 2 {"id": 2, "parent_id": 3} {"id": 3, "parent_id": 3}
|
19
|
+
# ...
|
20
|
+
#
|
21
|
+
# GroupConcat takes all the same options as Group.
|
22
|
+
#
|
23
|
+
# @see Group
|
24
|
+
class GroupConcat < Group
|
25
|
+
|
26
|
+
description <<EOF
|
27
|
+
This processor concatenates records of a consecutive group of records
|
28
|
+
into a single record.
|
29
|
+
|
30
|
+
$ cat input
|
31
|
+
{"id": 1, "parent_id": 4}
|
32
|
+
{"id": 2, "parent_id": 3}
|
33
|
+
{"id": 3, "parent_id": 3}
|
34
|
+
...
|
35
|
+
|
36
|
+
$ cat input | wu-local group_concat --by=parent_id --to=tsv
|
37
|
+
4 1 {"id": 1, "parent_id": 4}
|
38
|
+
3 2 {"id": 2, "parent_id": 3} {"id": 3, "parent_id": 3}
|
39
|
+
...
|
40
|
+
|
41
|
+
Each output record consists of tab-separated fields in the following
|
42
|
+
order:
|
43
|
+
|
44
|
+
1) The key defining the group of input records in this output record
|
45
|
+
2) The number of input records in the group
|
46
|
+
3) Each input record in the group
|
47
|
+
...
|
48
|
+
|
49
|
+
This processor will not produce any output for a given group until it
|
50
|
+
sees the last record of that group. See the documentation for the
|
51
|
+
'group' processor for more information.
|
52
|
+
EOF
|
53
|
+
|
54
|
+
# The members of the current group.
|
55
|
+
attr_accessor :members
|
56
|
+
|
57
|
+
# Initializes the empty members array.
|
58
|
+
def setup
|
59
|
+
super()
|
60
|
+
@members = []
|
61
|
+
end
|
62
|
+
|
63
|
+
# Initializes the empty members array.
|
64
|
+
#
|
65
|
+
# @param [Object] record
|
66
|
+
def start record
|
67
|
+
super(record)
|
68
|
+
self.members = []
|
69
|
+
end
|
70
|
+
|
71
|
+
# Accumulate each record, adding it to the current members.
|
72
|
+
#
|
73
|
+
# @param [Object] record
|
74
|
+
def accumulate record
|
75
|
+
super(record)
|
76
|
+
self.members << record
|
77
|
+
end
|
78
|
+
|
79
|
+
# Yields the group, including its key, its size, and each
|
80
|
+
# member.
|
81
|
+
#
|
82
|
+
# @yield [key, size, *members]
|
83
|
+
# @yieldparam [Object] key the key defining the group
|
84
|
+
# @yieldparam [Integer] size the number of members in the group
|
85
|
+
# @yieldparam [Array<Object>] the members of the group
|
86
|
+
def finalize
|
87
|
+
group = [key, size]
|
88
|
+
group.concat(members)
|
89
|
+
yield group
|
90
|
+
end
|
91
|
+
|
92
|
+
register
|
93
|
+
end
|
94
|
+
end
|
95
|
+
end
|
96
|
+
|
97
|
+
|
98
|
+
|
@@ -0,0 +1,71 @@
|
|
1
|
+
module Wukong
|
2
|
+
class Processor
|
3
|
+
|
4
|
+
# A base widget for building more complex improver widgets.
|
5
|
+
class Improver < Processor
|
6
|
+
|
7
|
+
# The current group of records.
|
8
|
+
attr_accessor :group
|
9
|
+
|
10
|
+
# Sets up this improver by defining an initial key (with a
|
11
|
+
# value that is unlikely to be found in real data) and calling
|
12
|
+
# `#zero` with no record.
|
13
|
+
def setup
|
14
|
+
@key = :__first_group__
|
15
|
+
zero
|
16
|
+
end
|
17
|
+
|
18
|
+
def recordize record
|
19
|
+
record.split("\t")
|
20
|
+
end
|
21
|
+
|
22
|
+
#
|
23
|
+
# All kinds of assumptions here,
|
24
|
+
# record is tab-delimited and the
|
25
|
+
# first field is a name of a function
|
26
|
+
# to call
|
27
|
+
#
|
28
|
+
def get_function record
|
29
|
+
record.first
|
30
|
+
end
|
31
|
+
|
32
|
+
# Processes the `record`.
|
33
|
+
def process(record)
|
34
|
+
fields = recordize(record)
|
35
|
+
func = get_function(fields)
|
36
|
+
case func
|
37
|
+
when 'zero' then
|
38
|
+
yield zero
|
39
|
+
when 'accumulate' then
|
40
|
+
accumulate(fields[1..-1])
|
41
|
+
when 'improve' then
|
42
|
+
yield improve(fields[1], self.group)
|
43
|
+
self.group = []
|
44
|
+
else
|
45
|
+
raise NoMethodError, "undefined method #{func} for Improver"
|
46
|
+
end
|
47
|
+
STDOUT.flush # WHY? Because.
|
48
|
+
end
|
49
|
+
|
50
|
+
# Starts accumulation for a new key. Return what you would
|
51
|
+
# with no improvements.
|
52
|
+
def zero
|
53
|
+
self.group = []
|
54
|
+
end
|
55
|
+
|
56
|
+
# Accumulates another +record+.
|
57
|
+
#
|
58
|
+
# @param [Object] record
|
59
|
+
def accumulate record
|
60
|
+
self.group << record
|
61
|
+
end
|
62
|
+
|
63
|
+
# Improve prev with group
|
64
|
+
#
|
65
|
+
#
|
66
|
+
def improve prev, group
|
67
|
+
end
|
68
|
+
|
69
|
+
end
|
70
|
+
end
|
71
|
+
end
|
@@ -0,0 +1,37 @@
|
|
1
|
+
module Wukong
|
2
|
+
class Processor
|
3
|
+
|
4
|
+
# Joins XML input data based on a root tag.
|
5
|
+
class JoinXML < Processor
|
6
|
+
|
7
|
+
field :root, String, default: 'xml', doc: "Name of the root XML element"
|
8
|
+
|
9
|
+
def setup
|
10
|
+
@lines = []
|
11
|
+
end
|
12
|
+
|
13
|
+
def process line
|
14
|
+
if match = terminator.match(line)
|
15
|
+
if match.end(0) == line.size
|
16
|
+
@lines << line
|
17
|
+
else
|
18
|
+
@lines << line[0...match.end(0)]
|
19
|
+
end
|
20
|
+
yield @lines.join("\n")
|
21
|
+
@lines = []
|
22
|
+
@lines << line[match.end(0)..-1] unless match.end(0) == line.size
|
23
|
+
else
|
24
|
+
@lines << line
|
25
|
+
end
|
26
|
+
end
|
27
|
+
|
28
|
+
def terminator
|
29
|
+
%r{<\s*/\s*#{root}\s*>}i
|
30
|
+
end
|
31
|
+
|
32
|
+
register :join_xml
|
33
|
+
end
|
34
|
+
end
|
35
|
+
end
|
36
|
+
|
37
|
+
|
@@ -0,0 +1,72 @@
|
|
1
|
+
require_relative("group")
|
2
|
+
|
3
|
+
module Wukong
|
4
|
+
class Processor
|
5
|
+
class Moments < Group
|
6
|
+
|
7
|
+
field :group_by, Whatever, :doc => "Part of the record to group by"
|
8
|
+
|
9
|
+
attr_accessor :measurements
|
10
|
+
|
11
|
+
field :of, Array, :default => [], :doc => "Parts of the record to measure moments of"
|
12
|
+
field :no_std_dev, :boolean, :doc => "Don't compute standard deviations"
|
13
|
+
|
14
|
+
def get_key record
|
15
|
+
super(record) unless (self.group_by || self.by)
|
16
|
+
get(self.group_by || self.by, record)
|
17
|
+
end
|
18
|
+
|
19
|
+
def receive_of o
|
20
|
+
@of = case o
|
21
|
+
when String then o.split(',')
|
22
|
+
when Array then o
|
23
|
+
else []
|
24
|
+
end
|
25
|
+
end
|
26
|
+
|
27
|
+
def start record
|
28
|
+
super(record)
|
29
|
+
@measurements = {}.tap do |m|
|
30
|
+
self.of.each do |property|
|
31
|
+
m[property] = []
|
32
|
+
end
|
33
|
+
end
|
34
|
+
end
|
35
|
+
|
36
|
+
def accumulate record
|
37
|
+
super(record)
|
38
|
+
self.of.each do |property|
|
39
|
+
if raw = get(property, record)
|
40
|
+
self.measurements[property] << (raw.to_f rescue next)
|
41
|
+
end
|
42
|
+
end
|
43
|
+
end
|
44
|
+
|
45
|
+
def results
|
46
|
+
{}.tap do |r|
|
47
|
+
measurements.each_pair do |property, values|
|
48
|
+
r[property] = {}
|
49
|
+
next if values.empty?
|
50
|
+
count = values.size.to_f
|
51
|
+
r[property][:count] = count.to_i
|
52
|
+
|
53
|
+
mean = values.inject(0.0) { |sum, value| sum += value } / count
|
54
|
+
r[property][:mean] = mean
|
55
|
+
unless no_std_dev
|
56
|
+
variance = values.inject(0.0) { |sum, value| diff = (value - mean) ; sum += diff * diff } / count
|
57
|
+
std = Math.sqrt(variance)
|
58
|
+
r[property][:std_dev] = std
|
59
|
+
end
|
60
|
+
end
|
61
|
+
end
|
62
|
+
end
|
63
|
+
|
64
|
+
def finalize
|
65
|
+
yield({:group => key, :count => size}.merge(:results => results))
|
66
|
+
end
|
67
|
+
|
68
|
+
register
|
69
|
+
end
|
70
|
+
end
|
71
|
+
end
|
72
|
+
|