WARNING: This project is no longer maintained in favour of PYLON Exporter ++.

PYLON-exporter

Utility for exporting data from a PYLON index. PYLON-exporter mandates a config driven approach to data collection, rather than code. It is the goal of this utility to support any combination of data extraction from a PYLON index.

Features:

• Simplified config driven approach abstracts complexities of data collection and merging
• Inbuilt queue to support large numbers of requests
• Parallel requests limit to manage control flow
• Custom result set merging
• Nested requests - native and custom
• Export as JSON or CSV

Setup

If Node is not installed, install it from https://nodejs.org/. Once complete, check by running node -v from a terminal window to show the install version.

Clone the repo:

git clone https://github.com/datasift/SE-PYLON-exporter.git

cd SE-PYLON-exporter

Install node module dependencies:

sudo npm install

Run tests:

npm test

Usage

node app.js

Quick Start

• Edit /config/basic.js and set authentication and index credentials.
• Run node app.js. This will load and execute the basic.js config file as default.
• Edit /config/basic.js as required.

For an example of all available config options see /config/all.js. To use this config file or any other, see Config Recipes section below.

Config Options

Below is a summary of all supported config options.

Option	Scope	Description
app.format	global	Output file format. csv or json
app.write_to_file	global	Boolean option to output data to file. true or false
app.max_parallel_tasks	global	The number of tasks to run in parallel.
app.log_level	global	Output log level. debug shows full requests and responses. info, warn, debug
auth.api_key	global	The api key used for authentication
auth.username	global	The username used for authentication
end	global, task	end time - unix timestamp. Defaults to now UTC
filter	global,task	A CSDL filter to apply to all tasks
id	global	The recording id of the index to analyze
id	merged tasks	A unique identifier for each merged task result set
name	task	A short, human readable description of the result set
start	global, task	start time - unix timestamp. Defaults to now -30 days UTC
then	freqDist task	Used to specify custom nested task properties
then.type	task	Override custom nested task types

Authentication

todo. See /config/basic.js

Output Format

JSON and CSV output formats are supported with JSON as default. Configure the output format by setting a. app.format property within the configuration file:

"app": {
    "format": "json" // json, csv
    }

Output Files

Output data can be exported to disk by setting an an app.write_to_file property. All files are written to the /output directory with a sub directory being created using the config file name e/g/ /output/test:

"app": {
    "write_to_file": false // true, false
    }

Config Recipes

A JSON configuration file is used to specify what data to extract from a PYLON index.

All config files inherit (and can overwite) properties from /config/default.js. This file is useful to place configurations you expect to use regularly e.g. primary authentication credentials.

Config File Usage

To specify which config file to use, set the NODE_ENV environment variable:

export NODE_ENV=foo

If NODE_ENV is not specified, the basic config file will be used i.e.load the /config/basic.js config file.

Example Config Recipe

To create a new config file, simply create a new .js file within the /config/ directory e.g. foo.js. This file must export an object with an analysis object containing either a freqDist or timeSeries objects or both.

 module.exports = {
     "analysis": {
         "freqDist": [
             {
                 "target": "fb.author.gender",
                 "threshold": 2
             }
         ],
         "timeSeries": [
             {
                 "interval": "week",
                 "span": 1
             }
         ]
     }
 };

When run, the default config file will be loaded, followed by the foo config file. Foo will overwrite any duplicate values within the default file.

Task Filters

Filter All Tasks

A filter can be supplied at the global level. If a global filter property is specified, it will apply across all tasks.

"filter": "fb.type == \"story\"",
"analysis": {
    "freqDist": [
        ....

Filter per Task

The filter parameter is supported as expected.

{
    "filter": "fb.sentiment == \"negative\"",
    "target": "fb.topics.name",
    "threshold": 2
}

If a global filter property is specified, it will apply across all tasks and will be joined with an AND to any filters specified in individual tasks. The above filters in the same config file would yield: would yield a filter of:

  "filter": "(fb.type == \"story\") and (fb.sentiment == \"negative\")"

Task Names

Task names are used to provide a short, human readable description of the result set being generated which is included as part of the output. If files are being generated, the name property is also used as the file name.

{
    "name": "freqDist_topics_sentiment_negative", //<-- task name
    "filter": "fb.sentiment == \"negative\"",
    "target": "fb.topics.name",
    "threshold": 2
}

Task Time Ranges

As default, the exporter will use a 32 day time range for all tasks. This can be overridden for all tasks or on a per task basis. Simply specify a start and/or end parameter as a unix timestamp.

As default end will be set to now and start set to now -32 days. Either parameter may be omitted.

Override All Tasks

To override time ranges for all tasks, specify a start and/or end parameter as a root key of the config file:

"start": <UNIX_TS>,
"end":  <UNIX_TS>,
"analysis": {
    "freqDist": [
        ....

Override Per Task

For individual and nested tasks (both native and custom), simply specify as part of the task. Custom nested tasks will automatically inherit the parent task values.

{
    "start": <UNIX_TS>,
    "end":  <UNIX_TS>,
    "target": "fb.author.gender",
    "threshold": 2,
    "then": {
        "target": "fb.topics.name",
        "threshold": 2,
    }
}

Merging Result Sets

Merging of response data is supported for both timeSeries and freqDist requests by grouping the request tasks to be merged as an array. A task name property is specified as the merged tasks array key and the task array itself as the value:

{
    "timeSeries_brands_by_week": [ //<-- task name
        {
            "id": "ford", //<-- task id
            "filter": "fb.content contains \"ford\"",
            "interval": "week",
            "span": 2
        },
        {
            "id": "honda",
            "filter": "fb.content contains \"honda\"",
            "interval": "week",
            "span": 2
        }
    ]
}

Task ID

An id property can be specified to identify each data set within a merged task with a human readable name. For example, using the sample configuration above where id properties have been specified as ford and honda would return the following JSON response:

 {
      "ford": [
          {
              "key": "25-34",
              "interactions": 565500,
              "unique_authors": 432800
          },
          ....
      ],
      "honda": [
          {
              "key": "25-34",
              "interactions": 366500,
              "unique_authors": 296600
          },
          ....
      ]
  }

If an id is not specified, the exporter will attempt to automatically generate one. The target property will be used as default if present. If the target is not specified (e.g. time series) or is a duplicate, the filter property will be used. If the filter is a duplicate a concatenation of the target and filter property is used.

Nested Tasks

Two types of nested tasks are supported: native nested and custom nested tasks:

Native Nested Tasks

Native nested tasks are supported using the same simplified format using a child object:

"freqDist": [
    {
        "name": "example_native_nested",
        "target": "fb.author.gender",
        "threshold": 2,
        "child": {
            "target": "fb.author.age",
            "threshold": 2,
            "child": {
                "target": "fb.media_type",
                "threshold": 2
            }
        }
    }
]    

Custom Nested Tasks

Custom nested tasks offer increased flexibility over native nested tasks by adding support for all targets (native nested tasks are currently restricted to low cardinality targets only). In addition, the filter property is supported meaning secondary tasks can query a different data set than the primary if required. The task type can also be overwritten to support nesting timeSeries within freqDist (see "Type Property" section below).

The workflow for custom nested tasks is simple in that each result key from a primary task is used to generates subsequent secondary tasks by using the key as a filter parameter.

Custom nested tasks are configured within the config file using the then object:

"freqDist": [
    {
        "name": "freqDist_topics_by_gender",
        "target": "fb.author.gender",
        "threshold": 2,
        "then": {
            "target": "fb.topics.name",
            "threshold": 5
        }
    }
]

The above example would generate a primary request using the fb.author.gender target. With each of the returned keys i.e. male and female, secondary requests would be made using the fb.topics.name target. This is done by auto generating the filter parameters for the secondary tasks e.g.:

"filter": "(fb.author.gender ==\"female\")"

Filter Property

The filter property is supported as part of both a primary, secondary or both tasks. Currently these filters operate independently i.e. specifying a filter in the primary task does not in any way get applied to secondary tasks.

{
    "filter": "fb.sentiment == \"positive\"",
    "target": "links.domain",
    "threshold": 5,
    "then": {
        "target": "fb.gender",
        "threshold": 2
    }
}

Specifying a filter as part of a secondary task results in the secondary filter being automatically appended to the primary result keys using an AND operator. For example:

{
    "target": "fb.type",
    "threshold": 4,
    "then": {
        "filter": "fb.author.gender == \"male\"",
        "target": "fb.author.age",
        "threshold": 2
    }
}

This would generate the following request for each fb.type key:

  "filter": "(fb.author.gender == \"male\") AND fb.type ==\"like\""

Of course both primary and secondary filter parameters can be used together. For example, if we wanted results for Male authors only for both the primary request and secondary:

{
    "target": "fb.type",
    "threshold": 4,
    "filter": "fb.author.gender == \"male\"",
    "then": {
        "target": "fb.topics.name",
        "threshold": 2
        "filter": "fb.author.gender == \"male\"",
    }
}

If a filter parameter is specified at the global level, it will be joined with any other filters using an AND. For example a global filter of:

  "filter": "fb.author.country_code == \"US\""

combined with task of:

{
    "filter": "fb.sentiment == \"negative\"",
    "target": "fb.topics.name",
    "threshold": 2
}

would yield a filter of:

  "filter": "(fb.author.country_code == \"US\") and (fb.sentiment == \"negative\")"

Type Property

It is possible to override the type of custom nested tasks for both freqDist and timeSeries by specifying a type property as part of the then object. This is powerful in that request types can be be mixed.

For example, you may wish to generate a timeSeries for each member of a tag tree. This could be done by specifying a timeSeries tasks for each with a unique filter property however this could take many lines of config depending on how many tags were within the tree. The simpler method is to do this dynamically by overriding the type from freqDist to timeSeries:

"freqDist": [
    {
        "target": "interaction.tag_tree.automotive.brand",
        "threshold": 5,
        "then": {
            "type": "timeSeries", //<-- override type
            "interval": "week" //<-- timeSeries properties
        }
    }
]

Likewise, the reverse is possible, from timeSeries to freqDist. For example, for each week, show the top topics:

"timeSeries": [
    {
        "interval": "week",
        "then": {
            "type": "freqDist", //<-- override type
            "target": "fb.topics.name", //<-- freqDist properties
            "threshold": 10
        }
    }
]

NOTE: Due to the fact individual requests are used, custom nested requests can be more susceptible to redaction i.e. each individual request must have an audience size of > 1000 unique authors.

Tableau Workbook

An example Tableau workbook is provided and accompanying config recipe (/config/standard-tableau.js). This config recipe dynamically builds a Tableau dashboard based on a specified pre-defined tag tree. The export will be most effective for high level tags like products or brands etc as each tag is drilled in to for demographics, topics and hashtags etc.

• Edit the (/config/standard-tableau.js). As default this recipe will look to use a tag named interaction.tag_tree.standard. It is best practice to use this naming convention for CSDL where you intend to use this export. If you have not used this tag naming convention, simply find and replace interaction.tag_tree.standard with the desired tag name.
• Set the exporter to use the standard-tableau config recipe: export NODE_ENV=standard-tableau
• Run the app: node app.js
• The Tableau workbook will be output in the same dir as the output files.

Source File Locations

At the time of writing, Tableau partially supports relative source file paths (meaning the generated workbook and source .csv files can be moved once created) however once the notebook is saved, the source file location paths are then hard coded to the workbook. If the source files or the workbook are then subsequently moved outside of the original dir where the workbook was first saved, the source file paths will need to be updated within the workbook. Simply edit the standard-tableau.tbw workbook using a text edit (its an XML file) and find/replace the directory=' locations.

Custom Tableau Recipes

To create a custom Tableau config recipe simply copy the example (/config/standard-tableau.js) and rename accordingly. Providing the config file name contains the string 'tableau', the default workbook will be output with the source files.

Feature Requests and Bugs

https://github.com/haganbt/PYLON-exporter/issues