Use PrettyTable library for human-readable table output by Muteeb-Haider · Pull Request #4994 · Ericsson/codechecker

Muteeb-Haider · 2026-07-29T09:05:20Z

Replace the custom to_table() implementation in twodim.py with the PrettyTable library (prettytable~=3.0). The old implementation produced tables that fell apart on long lines. PrettyTable handles long content gracefully with proper box-drawing borders and correct column alignment.

Rewrite to_table() using PrettyTable with SINGLE_BORDER style
Left-align all columns to match previous formatting convention
Support separate_head and headerless table modes
Add prettytable~=3.0 to analyzer/requirements.txt
Add prettytable~=3.0 to report-converter dev requirements

Other output formats (csv, json, rows, dictlist) are unchanged.

bruntib

Thanks for the improvement. It's nice that we can use an existing library for table creation.
The main purpose of this development would be to make a table fit to the screen. For example, if I store the reports to the server and query them by the following command, then the table breaks to multiple lines if the columns are too wide:

CodeChecker cnd results <run name>

I'm not sure if this library is able to detect the screen width. Also, the proportion of the columns could be set reasonably. For example, the "severity" column is always short. But we shouldn't hard-code the column types. But we could compute the maximum width of each column and the long ones could be shortened somehow.

Thank you!

bruntib

This is a really nice work, thank you! I have a few minor comments. After fixing those we can merge this PR.

Side note: I'm just thinking about execution in Jenkins. That is a popular platform for CodeChecker integration. In that environment there is no "terminal width", so it defaults to 80. It might make it difficult to visualize big tables there. Users can set the COLUMNS environment variable as a workaround, but I'm thinking of detecting Jenkins based on Jenkins-specific environment variables as some heuristics. It shouldn't be done in this PR, I'm just writing this note for the future.

bruntib · 2026-07-31T21:45:24Z

+    # variable first (allows the user to override), then fall back to the
+    # OS-reported terminal size, and finally to 80 columns when stdout is
+    # redirected to a file or pipe.
+    import os  # pylint: disable=import-outside-toplevel


Import libs at the top of the file.

bruntib · 2026-08-01T09:01:50Z

+    nat_widths = _compute_natural_widths(field_names, data_rows, show_header)
+    return _fit_table_to_width(table, nat_widths, terminal_width)


The _compute_natural_widths() should be called by _fit_table_to_width(), because nat_widths isn't used elsewhere.

bruntib · 2026-08-01T09:07:37Z

        # The statistics should only print the unique ones.
-        self.assertIn("Number of analyzer reports                       | 6",
-                      out)
+        self.assertIn("Number of analyzer reports", out)


Keep the number, because the goal of this test is to check the number.

Replace the custom to_table() implementation in twodim.py with the PrettyTable library (prettytable~=3.0). The old implementation produced tables that fell apart on long lines. PrettyTable handles long content gracefully with proper box-drawing borders and correct column alignment. - Rewrite to_table() using PrettyTable with SINGLE_BORDER style - Left-align all columns to match previous formatting convention - Support separate_head and headerless table modes - Add prettytable~=3.0 to analyzer/requirements.txt - Add prettytable~=3.0 to report-converter dev requirements Other output formats (csv, json, rows, dictlist) are unchanged.

pylint flagged separate_footer as an unused argument. Implement it properly: when True, the last row is printed as a separate table below the main one, visually distinguishing footer rows (e.g. totals) from regular data rows.

The analyze_and_parse functional tests compare CodeChecker parse output against stored .output files. Update all 61 expected output files to reflect the new PrettyTable table format (box-drawing borders) instead of the old plain-dash format.

pylint reported no-name-in-module because SINGLE_BORDER is a deprecated alias not listed in prettytable's __all__. Switch to the proper TableStyle.SINGLE_BORDER enum introduced in prettytable 3.x.

PrettyTable uses Unicode box-drawing characters (┌─┬─┐) which require UTF-8 encoding to render correctly. In CI environments with a non-UTF-8 locale, these characters were outputting replacement chars (?) causing golden-output mismatches in the analyze_and_parse functional tests. Set PYTHONIOENCODING, LC_ALL and LANG to UTF-8 in codechecker_env() so that all subprocess invocations produce stable UTF-8 output regardless of the runner's locale.

Two issues: 1. Leftover _supports_unicode() reference caused NameError - removed. 2. PrettyTable with SINGLE_BORDER adds a separator (├───┤) between rows only when hrules includes ROWS. The Summary table (separate_head=False) was rendering with a mid-row separator that doesn't belong there. Updated all 61 .output test files to match the correct output where the summary block has no separator between its two data rows.

The conversion script stripped the ├───┤ separator between the header row and data rows in headed tables. Restore it in all 61 .output files to match what PrettyTable SINGLE_BORDER actually produces.

…le format

- twodim.py: add _fit_table_to_width() which detects terminal width via shutil.get_terminal_size() and proportionally distributes column widths. Short columns keep their natural width; only wide columns are shortened. No column types are hard-coded — widths are computed from actual data. max_table_width is applied as a safety net for rounding errors. - test_analyze.py: restore the check that the processed-file count is actually 1 (not just that the label appears). The old assertIn with '| 1' broke because PrettyTable uses Unicode box-drawing characters instead of ASCII pipes, so the assertion now checks that the label and the token '1' appear on the same output line.

- Split overlong comment line (E501/C0301) - Rename MIN_COL_WIDTH to min_col_width (C0103 snake_case) - Refactor _fit_table_to_width() from 6 to 3 args by extracting natural-width computation into a separate _compute_natural_widths() helper, eliminating the too-many-positional-arguments warning (R0917)

- Change assigned from List[Optional[int]] to List[int] (initialised to 0) so mypy knows every element is always an int before use - Add type: ignore[assignment] on table.align = 'l' to suppress mypy complaint about PrettyTable's dict-typed align property

shutil.get_terminal_size() reads directly from the OS TTY and ignores the COLUMNS environment variable when a real TTY is attached. Check COLUMNS first so users can override the detected width, then fall back to shutil.get_terminal_size() and finally to 80 columns.

The previous proportional-share approach could truncate short columns like 'Severity' (8 chars) when the proportional share fell just below the natural width due to rounding. New approach: in each fixpoint iteration, lock any column whose natural width fits within the *equal share* (remaining / num_unassigned). This guarantees short columns always keep their full natural width. Only the genuinely wide columns are then shortened proportionally from what remains.

The reviewer asked to keep the '| 1' check to assert the processed-file count is exactly 1. PrettyTable uses │ (U+2502) instead of ASCII |, so replace the previous any()-loop with a direct assertIn using the Unicode box-drawing character.

- twodim.py: move 'import os' to top of file (was inline inside to_table) - twodim.py: move _compute_natural_widths() call inside _fit_table_to_width() so nat_widths is no longer an external parameter at the call site - test_store.py: restore the '| 6' check to assert the report count is exactly 6; use Unicode │ (U+2502) to match PrettyTable's separator

The PrettyTable output may have variable spacing around the number depending on column width calculations. Use assertRegex instead of assertIn to allow flexible whitespace around the value and separator.

PrettyTable pads column content with spaces, so the output is: '│ Number of analyzer reports │ 6 │' Use assertRegex with \s+ to match the whitespace padding while still asserting the count is exactly 6.

Clean up test assertions

bruntib

Thank you for the development! It's much more convenient to read this output.

Muteeb-Haider requested a review from bruntib as a code owner July 29, 2026 09:05

bruntib requested changes Jul 30, 2026

View reviewed changes

Comment thread analyzer/tests/functional/analyze/test_analyze.py Outdated

Muteeb-Haider requested a review from bruntib July 31, 2026 17:21

bruntib requested changes Aug 1, 2026

View reviewed changes

Muteeb-Haider added 20 commits August 1, 2026 14:40

Fix unused separate_footer argument in to_table()

2c9711f

pylint flagged separate_footer as an unused argument. Implement it properly: when True, the last row is printed as a separate table below the main one, visually distinguishing footer rows (e.g. totals) from regular data rows.

Fix: use TableStyle enum instead of deprecated SINGLE_BORDER constant

6c28d1a

pylint reported no-name-in-module because SINGLE_BORDER is a deprecated alias not listed in prettytable's __all__. Switch to the proper TableStyle.SINGLE_BORDER enum introduced in prettytable 3.x.

Fix test output files: restore missing header separator line

c6f20e7

The conversion script stripped the ├───┤ separator between the header row and data rows in headed tables. Restore it in all 61 .output files to match what PrettyTable SINGLE_BORDER actually produces.

Fix unsupported item assignment on PrettyTable align property

04327e8

Fix Summary table separator row and PrettyTable HRuleStyle import

ce9c6d3

Fix line too long in statistics.py

c74bc49

Fix test assertions: update old pipe-format table checks to PrettyTab…

968508d

…le format

Fix test_store_stats: use regex for flexible whitespace matching

353f403

The PrettyTable output may have variable spacing around the number depending on column width calculations. Use assertRegex instead of assertIn to allow flexible whitespace around the value and separator.

Fix test_store_stats: use regex with \s+ to match PrettyTable padding

286a57a

PrettyTable pads column content with spaces, so the output is: '│ Number of analyzer reports │ 6 │' Use assertRegex with \s+ to match the whitespace padding while still asserting the count is exactly 6.

Muteeb-Haider force-pushed the muteeb/prettytable-output branch 2 times, most recently from 4f5e8e1 to 959c24f Compare August 1, 2026 13:01

Update test assertions for PrettyTable format

b4fe95b

Clean up test assertions

Muteeb-Haider force-pushed the muteeb/prettytable-output branch from 959c24f to b4fe95b Compare August 1, 2026 13:02

Derive field names and rows from table object in width fitting

96e0428

bruntib approved these changes Aug 1, 2026

View reviewed changes

bruntib merged commit 91547e6 into Ericsson:master Aug 1, 2026
10 checks passed

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Uh oh!

Use PrettyTable library for human-readable table output - #4994

Use PrettyTable library for human-readable table output#4994
bruntib merged 22 commits into
Ericsson:masterfrom
Muteeb-Haider:muteeb/prettytable-output

Muteeb-Haider commented Jul 29, 2026

Uh oh!

bruntib left a comment

Uh oh!

Uh oh!

bruntib left a comment

Uh oh!

bruntib Jul 31, 2026

Uh oh!

bruntib Aug 1, 2026

Uh oh!

bruntib Aug 1, 2026

Uh oh!

bruntib left a comment

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

		nat_widths = _compute_natural_widths(field_names, data_rows, show_header)
		return _fit_table_to_width(table, nat_widths, terminal_width)

Uh oh!

Conversation

Muteeb-Haider commented Jul 29, 2026

Uh oh!

bruntib left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

bruntib left a comment

Choose a reason for hiding this comment

Uh oh!

bruntib Jul 31, 2026

Choose a reason for hiding this comment

Uh oh!

bruntib Aug 1, 2026

Choose a reason for hiding this comment

Uh oh!

bruntib Aug 1, 2026

Choose a reason for hiding this comment

Uh oh!

bruntib left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants