One name in sibling folders

This page is generated from the code and must not be edited by hand.

Files with exactly the same name in folders side by side under one folder (images/a.tif and masks/a.tif) are grouped together, unless the name turns up in so many side-by-side folders that it is just how each folder names its files.

one_name_in_sibling_folders version 1+d1

• Kind of rule: Which files go together (groups). Says which things count as the same in some respect, and so forms groups.

• What it concludes: equivalent, in respect of one_name_beside

• Reasoning: defeasible. A conclusion is usually right, and is retracted when a rule it names concludes against it.

• What it reads: files, which it sorts into groups

Is it available in this version?

Yes. The engine runs it.

What would defeat it

What this rule concludes about something is withdrawn when any of these rules concludes about the same thing:

• filesystem_noise_from_spelling: Files a computer makes on its own, such as .DS_Store or Thumbs.db, are recognized by how their names are spelled and set aside; an office program’s lock file, whose name starts ~$, is left alone, because it means somebody was at work.

• placeholder_from_entry_kind: A shortcut, or a file that claims a size but takes up no room on the disk, is taken to stand in for contents kept somewhere else.

• installed_software_in_folder: A folder is recognized as an installed program, such as a copy of Fiji or a Python environment, from the parts it holds, a name ending .app, its program-style subfolders or a majority of program files; its files are still indexed and read, and rules about experiments leave them alone.

It stands aside beneath every folder one of these rules raises:

• installed_software_in_folder

Parameters

Parameter	Shipped value	May be	What it does
maximum_folders	4	whole number, from 2 to 1000	The most sibling folders one name may lie in and form a group: a dataset’s images, labels, masks and predictions are four; a name in more is each sample’s own.
apart_folder_senses	["cross_validation", "control", "treatment", "wild_type", "knockout"]	list of text	Two sibling folders whose names agree but for their numbers and words bound to these senses hold different things numbered alike, and no file of one is paired with a file of the other: a learning dataset’s train, test and validation splits, and an experiment’s conditions.
ignored_roles	["readme", "licence", "changelog", "notes", "metadata"]	list of text	The roles of a file that describes its folder rather than holding its data: a README in each of two sibling folders is each folder’s own (the evaluation generator’s pool).
source_folder_senses	["raw", "acquisition", "image", "input_data"]	list of text	The senses a folder’s words say it holds what came first by: raw, recorded, images, input.
product_folder_senses	["annotation", "annotated", "masked", "segmented", "reference_data", "analysis", "processed"]	list of text	The senses a folder’s words say it holds what was made from them by: labels, masks, segmentations, ground truth, output, results and processed.