manifest

from plexora.api import manifest

Every surface in Plexora -- a plugin's Requires, the edit page, the Open Project card's "Needs setup" badge, the CLI, the Python API -- asks some form of "does this project have X yet". This module is the one place that answers it, so every reader agrees on what "has a table" means.

This module tracks three states, not two:

• PRESENT -- the user answered this, and the answer is recorded.
• GUESSED -- there is a value, but Plexora worked it out. A guess that happens to be right is still a guess, so the first tool that depends on one shows it back once, for confirmation.
• MISSING -- nothing, and nothing to show back.

Plus NOT_APPLICABLE, for a question this project's format cannot be asked at all: an AnnData's x/y do not exist as columns, because the adapter builds them from a read spec, so offering two column selects for them is a form with no possible answer in it. That is a different thing from missing.

MembersCopy Anchor Link

summaryCopy Anchor Link

summary(project: Project) -> dict

The short form a project card or a CLI listing is drawn from.

Deliberately not the whole manifest: a page listing two hundred projects wants a handful of facts about each, not the full set of questions.

statusCopy Anchor Link

status(project: Project, key: str) -> dict

The state of one question about a project.

value is what a surface shows back -- a path, a column name, a list -- or None. For "table" on a source that cannot yet be read it is the dict {src, table, unresolved}, which is what the requirements modal prefills its data field from: the user already gave the path, so asking for it again is asking a question they already answered.

answeredCopy Anchor Link

answered(project: Project, key: str) -> bool

Whether the project holds a value for this question, guessed or given.

Says nothing about who supplied the value -- a predictor's guess counts as answered here the same as something the user typed in. status() is what tells the two apart.

open_questionsCopy Anchor Link

open_questions(project: Project, keys=None) -> list[str]

Which questions this project still cannot answer, in ask order.

A role question ("role:cell_id", and the rest) is withheld while the table itself is unanswered: asking which column holds the cell id before any columns exist is a question with no possible answer, and "table" already covers it.

needs_setupCopy Anchor Link

needs_setup(project: Project) -> bool

Whether this project is holding a question it cannot answer itself.

Only the one state deserves a badge on a card: a data file was named and something about it is still undecided, so the project silently opens as an image and the user has no other way to find out why. An image-only project is never flagged -- that is a complete, valid project, and the whole premise of "an image alone is enough".

KEYSCopy Anchor Link

KEYS = ('image', 'segmentation', 'table', 'features', 'markers', 'coordinates', 'role:cell_id', 'role:x', 'role:y', 'role:image_id', 'role:celltype', 'role:z', 'role:t')

Every question core can record an answer to, in the order a form should ask them: the files first, then what is inside them, then what the columns mean.

PRESENTCopy Anchor Link

PRESENT = 'present'

The user answered this question, and the answer is recorded.

GUESSEDCopy Anchor Link

GUESSED = 'guessed'

There is a value, but Plexora worked it out rather than being told. A guess that happens to be right is still a guess.

MISSINGCopy Anchor Link

MISSING = 'missing'

Nothing is recorded, and there is nothing to show back.

NOT_APPLICABLECopy Anchor Link

NOT_APPLICABLE = 'not_applicable'

This project's format cannot be asked this question at all -- a different thing from missing, and not a gap for the user to fill.