Add support for configuration profiles #395
Conversation
dlax
changed the title
|
From what I understand the configuration we can set in the file relates to the columns of the query display. But I guess other options would be nice to have to. Did I miss something ? (is it something you planned to add afterwards ?) |
|
From what I understand the configuration we can set in the file relate to the columns of the query display.
@kmoppel-cognite was looking for a way to hide sections of the header (instance, system, workers).
These are not configurable through configuration files at the moment,
indeed; but we can certainly add support for these later on.
|
|
Hey, glad to see some progress! Looks promising :) At the same time - do I understand correctly that currently there is still no "infra" for built-in profiles where configs could be read in some fallback mode from some "flags dictionary" in code? |
Sorry, I don't get your question; can you reformulate? |
dlax
force-pushed
the
profile
branch
2 times, most recently
from
02a00a8 to
f7a86c3
Compare
This version has reached its end-of-life last year, https://peps.python.org/pep-0537/#lifespan.
We use the syntax from PEP 585 and PEP 604 and rely on 'from __future__ import annotations' to modernize the code base. However, not all generic aliases (e.g. Dict, List) can be replaced before Python 3.9.
Rephrase their help text and move then to a dedicated options group.
We introduce UIHeader class to hold header-specific options. Along the way, many names (variables, functions) are changed in order to make things more explicit.
In the command-line parser, options related to header information have their default value set to None in order to distinguish when the option was not specified from when it got, as it would store False then. This makes it possible to have a precedence of command-line options over the configuration file (implemented in types.UIHeader.make()). The validation of configuration options, previously in UISection, is extracted into a BaseSectionMixin and reused in new HeaderSection.
SectionProxy.get*() would never actually raise NoOptionError because the 'fallback' argument defaults to None.
This stems from the idea discussed at https://github.com/dalibo/pg_activity/pull/387#issuecomment-1623178466.#396.
One thing I wonder is whether the option should be named
--config <full-path-to-config>or if we should introduce another option later on.Based on #404