Change how Truncation works

[Trenly]

• I have signed the Contributor License Agreement.
• I have updated the Release Notes.
• This pull request is related to an issue.
  • Several, will link them after review and before merge
  • Resolves winget list truncation is not uniform #6137
  • Resolves avoiding truncating search/list result #1653
  • Resolves winget search output displays incomplete id when id is long #3986
  • Resolves Increasing terminal width doesn't address truncation #3071
  • Resolves flexible handling of GetConsoleWidth method's default value would be useful in scripting #1300
  • Resolves Package names are being truncated when searching with empty string #2603
  • Resolves Add '--autocolumnwidth' option #2161
  • Related to “winget list” don't get "Available" column #1618

Problem

When winget list output is redirected to a file or variable:

• Output was truncated at 120 characters (hardcoded fallback width)
• Rows beyond the first 50 were truncated to widths locked by earlier rows
• Spinner and progress bar characters appeared in the redirected output

Changes

ChannelStreams.h / ChannelStreams.cpp

• GetConsoleWidth() now returns std::optional<size_t> — nullopt when no console is attached (stdout redirected), actual width otherwise. Eliminates the hardcoded 120-column fallback.

TableOutput.h

• All rows are now buffered before any output is written. Column widths are computed from the full dataset in EvaluateAndFlushBuffer(), eliminating truncation and misalignment for both console and
  non-console output.
• Column shrinking to fit console width is skipped when no console is present.
• Removed the 50-row "sizing buffer" optimization that was the root cause of truncation even if the console was wide enough for all columns.

Note

I assumed the 50-row sizing buffer was for performance reasons and to ensure the CLI would not take too much memory. A test on my machine showed a terminal width of 220 and roughly 200 packages. Assuming a terminal width of 500 characters on an ultrawide monitor, UTF-16 encoded, with 10k rows, that would be roughly 10 MB of buffered content. Given that 10k rows and 500 characters would be a statistical anomaly, I assumed that it would be safe to remove this sizing buffer even if it comes at a minor performance cost.

ExecutionReporter.cpp

• Spinner and progress bar are only created when a console is present (GetConsoleWidth().has_value()). When stdout is redirected they remain nullptr, suppressing all progress noise from redirected output.

ExecutionProgress.cpp, WorkflowBase.cpp, ConfigurationFlow.cpp

• Updated call sites to handle std::optional<size_t> from GetConsoleWidth().

AppInstallerCLITests

• Added new test cases to ensure console width creates appropriate truncation behavior
• Added new test to ensure that buffering all lines at once does not cause issues

---

Microsoft Reviewers: Open in CodeFlow

[JohnMcPMS]

/azp run

[azure-pipelines]

Azure Pipelines successfully started running 1 pipeline(s).

[JohnMcPMS]

/azp run

[azure-pipelines]

Azure Pipelines successfully started running 1 pipeline(s).