pivot_forecast/readme.md

• setup postgres db with a username and password and database
• add this to .bashrc to quickly invoke command line connection export PGD="psql -U ptrowbridge -d ubm -p 5433 -h 192.168.1.110" setup

---

• setup the application tables $PGD -f setup_sql/schema.sql
• create a table of data to forecast
• edit target_info.sql to reflect the name of the table from previous step
• $PGD -f setup_sql/target_info.sql to populate the meta table
• fill out the target_meta table in the database to model the forecast table
• PGD -f setup_sql/build_maste_tables.sql to create master data tables from forecast data
• ./routes/baseline/generate_route_sql.sh to create the baseline sql used by the /baseline route

schema	tname	cname	opos	func	fkey	fcol	dtype	appcol	pretty
fc	live	fb_cst_loc	91	cost		fb_cst_loc	numeric
fc	live	ship_cust	36	scust	scust	ship_cust	text
fc	live	rdate	98	rdate	rdate	drange	date
fc	live	geo	42	scust		geo	text	customer
fc	live	part	54	item	item	part	text	item
fc	live	odate	96	odate	odate	drange	date	order_date
fc	live	sdate	100	sdate	sdate	sdate	date	ship_date
fc	live	oseas	97	odate		ssyr	integer
fc	live	calc_status	94	order_status	order_status	calc_status	text	order_status
fc	live	rseas	99	rdate		ssyr	integer
fc	live	sseas	101	sdate		ssyr	integer
								version
								iter
								logid

• func: table name of associated data
• fkey: primary key of assoicated dat
• fcol: associated field from the master data table if it is different (oseas would refer to ssyr in fc.perd)
• pretty: display column name (user friendly)
• appcol: parameters that will have to be supplied but the application
  • order_date
  • ship_date
  • customer
  • item
  • order_stats
  • units
  • cost
  • value
  • version (added if missing)
  • iter (added if missing)
  • logid (added if missing)

routes

• all routes would be tied to an underlying sql that builds the incremental rows
• that piece of sql will have to be build based on the particular sales layout
  • columns: a function to build the columns for each route
  • where a function to build the where clause will be required for each route
  • the result of above will get piped into a master function that build the final sql
  • the master function will need to be called to build the sql statements into files of the project

route baseline

• forecast = baseline (copied verbatim from actuals and increment the dates) + diffs. if orders are canceled this will show up as differ to baseline
• regular updates to baseline may be required to keep up with canceled/altered orders
• copy some period of actual sales and increment all the dates to serve as a baseline forecast

• join to period tables to populate season; requires variance number oof table joins, based on howmany date functions there are 🙄
• some of the app parameters can be consolidated, the baseline period could be one large range potentially, instead of 2 stacked periods
• setup something to fill in sql parameters to do testing on the function
• update node to handle forecast name parameter
• calc status is hard-coded right now in the json request -> probably needs to be manuall supplied up front

scale

• need to add version columns to all CTE's
• need to build log insert
• Need to build where clause for scenario

running problem list

• baseline route
  • problem: how will the incremented order season get updated, adding an interval won't work
    • a table fc.odate, has been built, but it is incomplete, a setup function filling in these date-keyed tables could be setup
    • if a table is date-keyed, fc.perd could be targeted to fill in the gaps by mapping the associated column names
  • problem: the target sales data has to map have concepts like order_date, and the application needs to know which col is order date
    • add column called application hook
  • there is not currently any initial grouping to limit excess data from all the document# scenarios
• general
  • clean up SQL generation to prevent injection
  • how to handle a target value adjustment, which currency is it in?
  • the sales data has to have a column for module and change ID, live sales data isn't going to work well
    • need to target the live sales data, build build a whole new table to use it plus add version columns