Tables: SizingFixedFit flag not honored correctly for synched tables (widths of one bleeds into next one)

[Ifry]

Version/Branch of Dear ImGui:

Version 1.90.1, Branch: master

Back-ends:

imgui_impl_dx11.cpp

Compiler, OS:

Windows 11 + MSVC 2022

Full config/build information:

No response

Details:

My Issue/Question:

When I use ImGuiTableFlags_SizingFixedFit without ImGuiTableFlags_Resizable, table columns are not synchronized and have different widths. It appears as if the widths are swapped between the first and second table.

The expected behavior is to have both tables with the same width of first column adjusted to fit the text "Very very long text 1"

Screenshots/Video:

Minimal, Complete and Verifiable Example code:

            ImGui::Begin("Synced tables test");

            if (ImGui::BeginTable("test_table", 2, ImGuiTableFlags_Borders | ImGuiTableFlags_SizingFixedFit))
            {
                ImGui::TableNextColumn();
                ImGui::Text("Very very long text 1");

                ImGui::TableNextColumn();
                ImGui::Text("Text 1");

                ImGui::TableNextColumn();
                ImGui::Text("Very very long text 2");

                ImGui::TableNextColumn();
                ImGui::Text("Text 2");

                ImGui::EndTable();
            }

            if (ImGui::BeginTable("test_table", 2, ImGuiTableFlags_Borders | ImGuiTableFlags_SizingFixedFit))
            {
                ImGui::TableNextColumn();
                ImGui::Text("Text 1");

                ImGui::TableNextColumn();
                ImGui::Text("Text 2");

                ImGui::TableNextColumn();
                ImGui::Text("Text 1");

                ImGui::TableNextColumn();
                ImGui::Text("Text 2");

                ImGui::EndTable();
            }

            ImGui::End();

[ScottDennison]

As a new user of ImGui, I thought I was going insane
But this exactly matches what I am seeing

[ocornut]

The issue here appears because they are tables sharing the same identifier.

If I change the code to use:

if (ImGui::BeginTable("test_table", 2, ImGuiTableFlags_Borders | ImGuiTableFlags_SizingFixedFit))
[...]
if (ImGui::BeginTable("test_table2", 2, ImGuiTableFlags_Borders | ImGuiTableFlags_SizingFixedFit))

Then it works as expected.

So effectively it is a bug in the feature that allows to share table settings between two tables.
I can totally investigate and aim to fix that but the correct result after a fix is that the first column of both tables will always be wide enough to hold the "Very very long text 1" bit, and I presume this isn't what you want. So in your specific situation I believe it would be more correct that you ensure your tables have different identifiers.

Read the FAQ Entry about ID Stack System if you need to submit many tables in a loop and you want them to have unique ids, then you can use e.g. PushID(loop_Index).

[Ifry]

the correct result after a fix is that the first column of both tables will always be wide enough to hold the "Very very long text 1" bit

This is exactly what i wanted

[ScottDennison]

I can't answer for OP, but at least in my case

the correct result after a fix is that the first column of both tables will always be wide enough to hold the "Very very long text 1" bit, and I presume this isn't what you want

This is exactly what I wanted.
Edit: Heh, OP responding with the exact same sentence as I typed this

I am having the tables be children of treenodes (via TreeNodeEx with NoPush), with all tables being the same tree depth. The tables do not have borders (it just looks like columns of text), and I was aiming for the columns to have the same positioning.

[ocornut]

I have pushed a fix for this d3c3514
Will try to write a proper automated test for it later so I am keeping this open for now.

[ocornut]

Pushed test for this too:
ocornut/imgui_test_engine@e7cf491

[ScottDennison]

Unfortunately, having updated the codebase that I was experiencing this issue in to v1.90.5, the issue remains.

The only notable difference from OP's original issue, is for me, the short text is in the first table and the long text is in the second table. I'll see if simply reordering the rows from OP's MCVE is enough to reproduce it, and if so, report back here.

[ScottDennison]

In fact, I am seeing this in the imgui demo (which shows v1.90.5 at the top)

Note that when expanding synced instance 1, synced instance 0 gains the extra blank space (good), but instance 1 does not and has any 2 digit cells truncated.