Compare commits

..

2 Commits

Author SHA1 Message Date
Shadowfacts f84d849432
Add conditional stage
Allows applying another pipeline stage based on a condition, which can
either be a whole filter or a single filter rule.
2019-11-01 22:50:25 -04:00
Shadowfacts 13c44d5e10
Refactor filtering logic into separate module 2019-11-01 22:49:52 -04:00
3 changed files with 216 additions and 117 deletions

View File

@ -0,0 +1,87 @@
defmodule Frenzy.Pipeline.ConditionalStage do
require Logger
alias Frenzy.Pipeline.{Stage, FilterEngine}
@behaviour Stage
@filter_modes ["accept", "reject"]
@rule_modes ["contains_string", "matches_regex"]
@impl Stage
def apply(%{"stage" => stage, "opts" => stage_opts, "condition" => condition}, item_params) do
if test_condition(condition, item_params) do
apply(String.to_existing_atom("Elixir." <> stage), :apply, [stage_opts, item_params])
else
{:ok, item_params}
end
end
@impl Stage
def apply(opts, item_params) do
Logger.warn("Received invalid conditional opts: #{inspect(opts)}")
{:ok, item_params}
end
defp test_condition(%{"mode" => mode} = filter, item_params) when mode in @filter_modes do
FilterEngine.test(filter, item_params)
end
defp test_condition(%{"mode" => mode} = rule, item_params) when mode in @rule_modes do
FilterEngine.test_rule(rule, item_params)
end
defp test_condition(condition, _item_params) do
Logger.warn("Received invalid condition: #{inspect(condition)}")
false
end
@impl Stage
def validate_opts(opts) do
cond do
not (Map.has_key?(opts, "stage") and is_binary(opts["stage"]) and
module_exists(opts["stage"])) ->
{:error, "stage must be a string containg a module that exists"}
not (Map.has_key?(opts, "opts") and is_map(opts["opts"])) ->
{:error, "opts must be a map"}
not (Map.has_key?(opts, "condition") and is_map(opts["condition"])) ->
{:error, "condition must be a map"}
true ->
with {:ok, stage_opts} <-
apply(String.to_existing_atom("Elixir." <> opts["stage"]), :validate_opts, [
opts["opts"]
]),
{:ok, condition} <- validate_condition(opts["condition"]) do
{
:ok,
opts
|> Map.put("opts", stage_opts)
|> Map.put("condition", condition)
}
else
{:error, _reason} = err ->
err
end
end
end
defp module_exists(module_name) do
try do
String.to_existing_atom("Elixir." <> module_name)
true
rescue
ArgumentError -> false
end
end
defp validate_condition(%{"mode" => mode} = filter) when mode in @filter_modes do
FilterEngine.validate_filter(filter)
end
defp validate_condition(%{"mode" => mode} = rule) when mode in @rule_modes do
FilterEngine.validate_rule(rule)
end
defp validate_condition(_condition), do: {:error, "condition must be either a filter or a rule"}
end

View File

@ -0,0 +1,120 @@
defmodule Frenzy.Pipeline.FilterEngine do
def validate_filter(filter) when is_map(filter) do
cond do
not is_map(filter) ->
{:error, "filter must be a map"}
not (Map.has_key?(filter, "mode") and is_binary(filter["mode"]) and
filter["mode"] in ["accept", "reject"]) ->
{:error, "mode must be a string, either 'accept' or 'reject'"}
not (Map.has_key?(filter, "score") and is_integer(filter["score"])) ->
{:error, "score must be an integer"}
not (Map.has_key?(filter, "rules") and is_list(filter["rules"])) ->
{:error, "rules must be a list of rules"}
true ->
validate_rules(filter)
end
end
def validate_rules(%{"rules" => rules} = filter) do
case validate_rules(rules) do
{:ok, rules} ->
{:ok, Map.put(filter, "rules", rules)}
{:error, _reason} = err ->
err
end
end
def validate_rules(rules) when is_list(rules) do
rules
|> Enum.with_index()
|> Enum.reduce_while({:ok, []}, fn {rule, index}, {:ok, new_rules} ->
case validate_rule(rule) do
{:ok, rule} ->
{:cont, {:ok, [rule | new_rules]}}
{:error, reason} ->
{:halt, {:error, "invalid rule #{index}: #{reason}"}}
end
end)
end
def validate_rules(_rules), do: {:error, "rules must be a list"}
def validate_rule(rule) do
cond do
not is_map(rule) ->
{:error, "rule must be a map"}
not (Map.has_key?(rule, "mode") and is_binary(rule["mode"]) and
rule["mode"] in ["contains_string", "matches_regex"]) ->
{:error, "mode property must be a string, either 'contains_string' or 'matches_regex'"}
not (Map.has_key?(rule, "property") and is_binary(rule["property"]) and
rule["property"] in ["url", "title", "author"]) ->
{:error, "property property must be a string, either 'url', 'title', or 'author'"}
not (Map.has_key?(rule, "param") and is_binary(rule["param"])) ->
{:error, "param property must be a string"}
not (Map.has_key?(rule, "weight") and is_integer(rule["weight"])) ->
{:error, "weight property must be an integer"}
true ->
{:ok, rule}
end
end
def test(%{"mode" => mode, "score" => score, "rules" => rules}, item_params) do
item_score =
rules
|> Enum.filter(&test_rule(&1, item_params))
|> Enum.map(& &1["weight"])
|> Enum.sum()
matches = item_score >= score
case {mode, matches} do
{"accept", true} ->
true
{"reject", false} ->
true
_ ->
false
end
end
def test_rule(
%{"mode" => mode, "property" => property, "param" => param},
item_params
) do
with prop_value <- get_property(item_params, property),
true <- is_binary(prop_value),
true <- matches(prop_value, mode, param) do
true
else
_ ->
false
end
end
defp matches(value, "contains_string", param) do
String.contains?(value, param)
end
defp matches(value, "matches_regex", param) do
{:ok, regex} = Regex.compile(param)
String.match?(value, regex)
end
defp get_property(item_params, "url"), do: item_params.url
defp get_property(item_params, "title"), do: item_params.title
defp get_property(item_params, "author"), do: item_params.author
defp get_property(_item_params, _property), do: {:error, "invalid property"}
end

View File

@ -1,128 +1,20 @@
defmodule Frenzy.Pipeline.FilterStage do defmodule Frenzy.Pipeline.FilterStage do
require Logger require Logger
alias Frenzy.Pipeline.Stage alias Frenzy.Pipeline.{Stage, FilterEngine}
@behaviour Stage @behaviour Stage
@impl Stage @impl Stage
def apply(%{"mode" => mode, "score" => score, "rules" => rules}, item_params) def apply(filter, item_params) do
when is_binary(mode) and is_integer(score) and is_list(rules) do if FilterEngine.test(filter, item_params) do
item_score = {:ok, item_params}
rules
|> Enum.map(fn rule -> test(rule, item_params) end)
|> Enum.sum()
matches = item_score >= score
case {mode, matches} do
{"accept", true} ->
{:ok, item_params}
{"reject", false} ->
{:ok, item_params}
_ ->
Logger.debug("Skipping item #{item_params.url} due to feed filter")
:tombstone
end
end
@impl Stage
def apply(opts, item_params) do
Logger.warn("Received invalid filter opts: #{opts}")
{:ok, item_params}
end
@impl Stage
def validate_opts(opts) when is_map(opts) do
cond do
not (Map.has_key?(opts, "mode") and is_binary(opts["mode"]) and
opts["mode"] in ["accept", "reject"]) ->
{:error, "mode must be a string, either 'accept' or 'reject'"}
not (Map.has_key?(opts, "score") and is_integer(opts["score"])) ->
{:error, "score must be an integer"}
not (Map.has_key?(opts, "rules") and is_list(opts["rules"])) ->
{:error, "rules must be a list of rules"}
true ->
validate_rules(opts)
end
end
@impl Stage
def validate_opts(_opts), do: {:error, "options must be a map"}
defp validate_rules(%{"rules" => rules} = opts) do
rules
|> Enum.with_index()
|> Enum.reduce_while(:ok, fn {rule, index}, :ok ->
case validate_rule(rule) do
:ok ->
{:cont, :ok}
{:error, reason} ->
{:halt, {:error, "invalid rule #{index}: #{reason}"}}
end
end)
|> case do
:ok ->
{:ok, opts}
{:error, _reason} = err ->
err
end
end
defp validate_rule(rule) do
cond do
not is_map(rule) ->
{:error, "rule must be a map"}
not (Map.has_key?(rule, "mode") and is_binary(rule["mode"]) and
rule["mode"] in ["contains_string", "matches_regex"]) ->
{:error, "mode property must be a string, either 'contains_string' or 'matches_regex'"}
not (Map.has_key?(rule, "property") and is_binary(rule["property"]) and
rule["property"] in ["url", "title", "author"]) ->
{:error, "property property must be a string, either 'url', 'title', or 'author'"}
not (Map.has_key?(rule, "param") and is_binary(rule["param"])) ->
{:error, "param property must be a string"}
not (Map.has_key?(rule, "weight") and is_integer(rule["weight"])) ->
{:error, "weight property must be an integer"}
true ->
:ok
end
end
defp test(
%{"mode" => mode, "property" => property, "param" => param, "weight" => weight},
item_params
) do
with prop_value <- get_property(item_params, property),
true <- is_binary(prop_value),
true <- matches(prop_value, mode, param) do
weight
else else
_ -> Logger.debug("Skipping item #{item_params.url} due to feed filter")
0 :tombstone
end end
end end
def matches(value, "contains_string", param) do @impl Stage
String.contains?(value, param) def validate_opts(opts) do
FilterEngine.validate_filter(opts)
end end
def matches(value, "matches_regex", param) do
{:ok, regex} = Regex.compile(param)
String.match?(value, regex)
end
defp get_property(item_params, "url"), do: item_params.url
defp get_property(item_params, "title"), do: item_params.title
defp get_property(item_params, "author"), do: item_params.author
defp get_property(_item_params, _property), do: {:error, "invalid property"}
end end