zeutschler / tinyolap Goto Github PK

Rules need to be written in python. This requires 'real' Python coding and quite some boilerplate code. Sample:

@rule(cube="sales", trigger="delta %") def my_rule(c: Cell): return c["delta"] / c.plan["plan"]

Why not let users focus on the core rules itself and do the wrapping for them. This should not be too hard. sample:

["delta %"] = ["delta"] / ["plan"]

This would be a good bridge for users from TM/1 and alike.
However, this will only work with simple one-line rules.

Implement the good old "Splasher" to aid users in planning processes. Implement natural language (kind of) planning language.

from: register_rule(self, function, trigger: list[str] = None,
feeder: list[str] = None,
scope: RuleScope = None,
injection: RuleInjectionStrategy = None,
code: str = None):

to: register_rule(self, function, trigger: str | list[str] = None,
scope: RuleScope = None,
feeder: str | list[str] = None,
injection: RuleInjectionStrategy = None,
code: str = None):

Like TM/1 STET function.

^{Created from JetBrains using CodeStream}

The member class should represent a single member of a dimension, either context aware (bound to a cube) or stand alone. Purpose is to ease implementation of business rules e.g.

`sales, price, quantity = c.member("Measures, "Sales", "Price", "Quantity")

c[sales] = c[price] * c[quantity]`

All methods for Members, Subset, Attributes etc. should be moved to subclasses for more convenient and better structured access.

Data integration back and forth between Pandas data frames and TinyOlap.

Ideally one code base for both Views and FlexViews

• Manage state of individual open Views in TinyOlap, especially lifecycle management incl. ageing /automatic clean up, persistence, refresh policies etc.

• Drill down, rollup, drill-across along one or multiple dimensions.Support for frontends needed on what activities are available.

• Introduce new 'ViewBuilder' class to enable custom view creation and design, e.g. custom row and column layouts with alternating dimensions or overwrites of filter members.

.. automodule:: tinyolap.samples.tutor.load
:members:
:noindex:

see file samples.rst

member_idx = self.member_counter
this will fail (duplicate member index ) after members get deleted and added again.

[tinyolap] tinyolap/dimension.py (Line 217)

            member_idx = self.member_counter

Open in IDE · Open on GitHub

^{Created from JetBrains using CodeStream}

Implement Navigation Function for Member class: ±20 methods

[tinyolap] tinyolap/member.py (Line 58)

region Navigation functions

Open in IDE · Open on GitHub

^{Created from JetBrains using CodeStream}

Introduce tox for testing.

See also: https://christophergs.com/python/2020/04/12/python-tox-why-use-it-and-tutorial/

We need to get ride of the explicit measures. Instead

• the last dimension should automatically become the measure dimension.
  Question here: how to deal with aggregations?

or

• there should be no measures at all.

This is a big topic and requires a dedicated branch.

Implement and test for database snapshot generation and management.

Snapshot are required for database backup and time-travel capabilities

There should be a set of sample databases to showcase tiny-lap capabilities for different purposes.

Note: Databases should be delivered as code, so users can see how they've been created.

• All errors raised though rule execution should raise the same build-in rules errors.
• Users should be able to switch off error raising at database runtime to not crash the database if errors are not well handled.

We should have .register for external rules and .add for rules that should be added to the database.

^{Created from JetBrains using CodeStream}

Sample:

dim.add_member("Year", ("Q1", "Q2", "Q3", "Q4"))

dim.add_member(["Q1", "Q2", "Q3", "Q4"],
                   [("Jan", "Feb", "Mar"), ("Apr", "Mai", "Jun"),
                    ("Jul", "Aug", "Sep"), ("Oct", "Nov", "Dec")])

dim.add_member(["HY1", "HY2"], [("Jan", "Feb", "Mar", "Apr", "Mai", "Jun"),
                                    ("Jul", "Aug", "Sep", "Oct", "Nov", "Dec")])

The level of "Year" currently returns 1, but need to return 2,
as 2 is the depth of the resulting hierarchy, e.g., "Year"(2) > "Q1"(1) > "Jan"(0).

slots will provide some nice performance and memory consumption benefits we want to gain

'DataIntegrator' class to provide more simple capabilities to integrate data from external data sources for dimensions, attributes and cubes.

e.g. add_member( .... after:"this")
add_member( .... after:"this")
move_member(... after:"this", before:"that")

^{Created from JetBrains using CodeStream}

Create a set of 6 to 10 sample Views for the enterprise sample database. At least 2 to 3 per cube.