A scrolling list needs a precise answer about which rows belong in view. WindowSpan gives that index calculation a Windows coding workspace, with a compatible model connection and proposed changes you can review against boundary examples.
from math import isfinite, ceil
def row_window(count, row_height, offset, viewport, overscan=1):
if type(count) is not int or count < 0 or type(overscan) is not int or overscan < 0:
raise ValueError("rows.count")
if not all(type(x) in (int, float) and isfinite(x) for x in (row_height, offset, viewport)) or row_height <= 0 or viewport < 0:
raise ValueError("rows.geometry")
start_units = max(0, offset) / row_height
first = count if start_units >= count else int(start_units)
if first == count or viewport == 0:
return first, first
span = viewport / row_height
end = count if span >= count - start_units else min(count, ceil(start_units + span) + overscan)
return max(0, first - overscan), endfrom math import isfinite, ceil
def row_window(count, row_height, offset, viewport, overscan=1):
if type(count) is not int or count < 0 or type(overscan) is not int or overscan < 0:
raise ValueError("rows.count")
if not all(type(x) in (int, float) and isfinite(x) for x in (row_height, offset, viewport)) or row_height <= 0 or viewport < 0:
raise ValueError("rows.geometry")
start_units = max(0, offset) / row_height
first = count if start_units >= count else int(start_units)
if first == count or viewport == 0:
return first, first
span = viewport / row_height
end = count if span >= count - start_units else min(count, ceil(start_units + span) + overscan)
return max(0, first - overscan), endA scrolling list can have correct records and still show the wrong slice of them. WindowSpan provides a Windows AI coding workspace for investigating how interface geometry connects to data selection. Open the component's project and describe the scroll position, available space and rows that should appear. Ask the assistant to inspect both the range calculation and the code that consumes it.
The visible-row example is one development starting point. You might examine behavior after the data count shrinks, when the viewport has no height, or when scrolling reaches the end. Request changes only after deciding which boundaries your interface needs. The example is not a bundled virtual-list component, and its assumptions about row height must be checked against the application you are building.
For abliterated AI model choice, connect an available model through a supported OpenAI-compatible or Anthropic endpoint with credentials when required. Policies, context and tool access depend on the selected model and server. Review the range and rendering changes together, run project checks, and exercise scrolling and resizing in the actual interface. A calculation that returns plausible indexes still needs verification where those indexes become visible rows.
Provide the item count, row-height assumption, viewport size and scroll offset, along with the expected range. Include the rendering code that consumes the range. This separates a calculation problem from stale data or component rendering behavior.
Do not assume that a fixed-height calculation covers variable-height content. Describe the actual layout and ask for an approach that matches it. Inspect and test the proposal with representative row sizes before adopting it in your project.
No performance result is promised. The workspace helps you discuss and review implementation changes. Measure the real interface under comparable conditions and verify visible behavior; fewer rendered items alone do not establish an improvement in your application.
Check the endpoint's tool support and narrow the supplied context to relevant files. Capabilities depend on the model and server. A separately hosted local model does not automatically gain project access or make every connected tool local.
WindowSpan · Updated 2026-10-25
