tps/database
2018-12-02 00:59:00 -05:00
..
deploy change to match up with api 2018-06-28 01:01:02 -04:00
interface move everything into one folder 2018-06-25 00:01:05 -04:00
reports move everything into one folder 2018-06-25 00:01:05 -04:00
templates move everything into one folder 2018-06-25 00:01:05 -04:00
upgrade_scripts work on upgrade query to adjust schema and leave sources in place 2018-12-02 00:59:00 -05:00
build_maps.xlsm move everything into one folder 2018-06-25 00:01:05 -04:00
readme.md move everything into one folder 2018-06-25 00:01:05 -04:00

Generic Data Transformation Tool

The goal is to:

  1. house external data and prevent duplication on insert
  2. facilitate regular exression operations to extract meaningful data
  3. be able to reference it from outside sources (no action required) and maintain reference to original data

It is well suited for data from outside systems that

  • requires complex transformation (parsing and mapping)
  • original data is retained for reference
  • don't feel like writing a map-reduce

use cases:

  • on-going bank feeds
  • jumbled product lists
  • storing api results

The data is converted to json by the importing program and inserted to the database. Regex expressions are applied to specified json components and the results can be mapped to other values.

Major Interactions

  • Source Definitions (Maint/Inquire)
  • Regex Instructions (Maint/Inquire)
  • Cross Reference List (Maint/Inquire)
  • Run Import (Run Job)

Interaction Details

  • Source Definitions (Maint/Inquire)

    • display a list of existing sources with display detials/edit options

    • create new option

    • underlying function is tps.srce_set(_name text, _defn jsonb)

    • the current definition of a source includes data based on bad presumptions:

      • how to load from a csv file using COPY
      • setup a Postgres type to reflect the associated columns (if applicable)
  • Regex Instructions (Maint/Inquire)

    • display a list of existing instruction sets with display details/edit options
    • create new option
    • underlying function is tps.srce_map_def_set(_srce text, _map text, _defn jsonb, _seq int) which takes a source "code" and a json
  • Cross Reference List (Maint/Inquire)

    • first step is to populate a list of values returned from the instructions (choose all or unmapped) tps.report_unmapped(_srce text)
    • the list of rows facilitates additional named column(s) to be added which are used to assign values anytime the result occurs
    • function to set the values of the cross reference tps.srce_map_val_set_multi(_maps jsonb)
  • Run Import

    • underlying function is tps.srce_import(_path text, _srce text)

source definition

  • load data
    • the brwosers role is to extract the contents of a file and send them as a post body to the backend for processing under target function based on srce defintion
      • the backend builds a json array of all the rows to be added and sends as an argument to a database insert function
        • build constraint key based on srce definition
        • handle violations
        • increment global key list (this may not be possible depending on if a json with variable length arrays can be traversed)
        • build an import log
        • run maps (as opposed to relying on trigger)
  • read data
    • the schema key contains either a text element or a text array in curly braces
      • forcing everything to extract via #>{} would be cleaner but may be more expensive than jsonb_populate_record
      • it took 5.5 seconds to parse 1,000,000 rows of an identicle google distance matrix json to a 5 column temp table
    • top level key to table based on jsonb_populate_record extracting from tps.type developed from srce.defn->schema
    • custom function parsing contents based on #> operator and extracting from srce.defn->schema
    • view that uses the source definiton to extrapolate a table?
    • a materialized table is built based on the source definition and any addtional regex?
      • add regex = alter table add column with historic updates?
      • no primary key?
      • every document must work out to one row
{
    "name":"dcard",
    "source":"client_file",
    "loading_function":"csv"
    "constraint":[
        "{Trans. Date}",
        "{Post Date}"
    ],
    "schemas":{
        "default":[
            {
                "path":"{doc,origin_addresses,0}",
                "type":"text",
                "column_name":"origin_address"
            },
            {
                "path":"{doc,destination_addresses,0}",
                "type":"text",
                "column_name":"origin_address"
            },
            {
                "path":"{doc,status}",
                "type":"text",
                "column_name":"status"
            }
            {
                "path":"{doc,rows,0,elements,0,distance,value}",
                "type":"numeric",
                "column_name":"distance"
            }
            {
                "path":"{doc,rows,0,elements,0,duration,value}",
                "type":"numeric",
                "column_name":"duration"
            }
        ],
        "version2":[]
    }
}