Add automatic detection and merging of separate date/time column (#1220) #1259

lucabonamini · 2026-01-07T20:52:41Z

Summary

Added automatic detection of CSV files with date and time stored in separate columns, with automatic combination into a single timestamp for plotting.

Motivation

Some data logging equipment stores date and time information in separated CSV columns:

Date,Time,Temperature
2026-01-01,10:30:25.000,23.5
2026-01-01,10:30:26.000,23.6

PlotJuggler previously required timestamps in a single column.

Solution

Implemented automatic detection and merging:

New column types: Added DATE_ONLY and TIME_ONLY to ColumnType enum
Detection logic: Automatically identifies date-only columns (e.g., "2026-01-01") and time-only columns (e.g., "10:30:25.000")
Pairing algorithm: Finds adjacent date/time column pairs
Virtual columns: Creates combined "Date + Time" virtual columns in the UI
Timestamp generation: Parses and combines separate date/time values into Unix epoch timestamps with fractional second support

Timeline slider fix

The timeline slider was diving the time range by the number of data points instead of the number of intervals between them, causing navigation steps to display incorrect time values. This fix ensures the slider accurately represents the actual time spacing in the data.

Fix tooltip display at end of time series data

When the cursor is positioned at or beyond the last data point in a time series, the tooltip now correctly shows the final data point value instead of disappearing. This ensures consistent tooltip behavior across the entire data range, including at the boundaries.

lucabonamini · 2026-01-07T20:53:40Z

Related to #1220

facontidavide

first pass of suggested changes. also please remember to apply formatting with pre-commit before submitting a PR

facontidavide · 2026-01-10T17:39:43Z

plotjuggler_app/curve_tracker.cpp

+    else if (index >= curve->dataSize())
+    {
+      // Target is at or beyond the last point - return the last point
+      return curve->sample(curve->dataSize() - 1);
+    }


what is the purpose? also, do not include in the PR changes that are not related

I've included two small fixes as reported in the PR description (Fix tooltip display at end of time series data)

facontidavide · 2026-01-10T17:42:23Z

plotjuggler_plugins/DataLoadCSV/timestamp_parsing.cpp

+    for (const char* fmt : date_formats)
+    {
+      std::istringstream in{ base_str };
+      in.imbue(std::locale::classic());
+      date::year_month_day ymd;
+      in >> date::parse(fmt, ymd);
+      if (!in.fail())
+      {
+        info.type = ColumnType::DATE_ONLY;
+        info.format = fmt;
+        return info;
+      }
+    }


can you please use a lambda or a small helper function to avoid code repetition?

facontidavide · 2026-01-10T18:03:38Z

plotjuggler_plugins/DataLoadCSV/dataload_csv.cpp

+    }
+
+    // Find DATE_ONLY and TIME_ONLY column pairs
+    for (size_t date_idx = 0; date_idx < column_types.size(); date_idx++)


my suggestion is to limit to consecutive columns only, therefore change this to

Suggested change

for (size_t date_idx = 0; date_idx < column_types.size(); date_idx++)

using PJ::CSV::ColumnType;

for (size_t index = 0; (index + 1) < column_types.size(); index++)

{

const auto date_idx = index;

const auto tie_idx = index + 1;

if((column_types[date_idx].type!= ColumnType::DATE_ONLY) ||

(column_types[time_idx].type!= ColumnType::TIME_ONLY))

{

continue; // early return is easier to read

}

// is_adjacent is implicitly true

facontidavide · 2026-01-10T18:05:20Z

plotjuggler_plugins/DataLoadCSV/dataload_csv.cpp

+                combined.virtual_name = virtual_name;
+                _combined_columns.push_back(combined);
+
+                // Add to the UI list widget ONLY (not to column_names, which is used for CSV parsing)


I am not sure that I want to add a column in the UI. At most a not in a QLabel?

facontidavide · 2026-01-10T18:08:13Z

plotjuggler_plugins/DataLoadCSV/dataload_csv.cpp

+  if (lines.size() > 0)
+  {


I usually prefer early returns

Suggested change

if (lines.size() > 0)

{

if (lines.empty())

{

return;

}

facontidavide · 2026-01-10T18:16:12Z

I wonder if we shouldn't add a radio option for this case instead of relying on magically finding the pair and adding it to the columns.

facontidavide · 2026-01-10T18:26:02Z

plotjuggler_plugins/DataLoadCSV/timestamp_parsing.cpp

+      "%H:%M:%S",
+      "%H:%M",
+      "%I:%M:%S %p",  // 12-hour with AM/PM
+      "%I:%M %p"


I don't think we should accept times that don't have seconds

lucabonamini · 2026-01-10T20:12:55Z

plotjuggler_app/mainwindow.cpp

        min_time = std::min(min_time, t0);
        max_time = std::max(max_time, t1);
-        max_steps = std::max(max_steps, (int)data.size());
+        max_steps = std::max(max_steps, (int)data.size() - 1);


here the other fix (Timeline slider fix)

lucabonamini · 2026-01-10T20:58:48Z

I wonder if we shouldn't add a radio option for this case instead of relying on magically finding the pair and adding it to the columns.

Done. I've replaced the automatic list widget item with a dedicated radio button option. When adjacent date+time column pairs are detected, the radio button becomes enabled and shows which columns will be combined.

lucabonamini added 2 commits January 7, 2026 21:33

feat: add automatic detection and merging of separate date/time columns

0c6bec9

fix: timeline slider step calculation off-by-one error

7fc2532

fix: tooltip not appearing for last data point in timeseries

aaf59a0

facontidavide mentioned this pull request Jan 10, 2026

Inherit legend settings when splitting plot views #1260

Open

facontidavide requested changes Jan 10, 2026

View reviewed changes

facontidavide reviewed Jan 10, 2026

View reviewed changes

lucabonamini commented Jan 10, 2026

View reviewed changes

Address PR review feedback for CSV date/time column handling

0a7615b

lucabonamini requested a review from facontidavide January 10, 2026 20:58

-    for (size_t date_idx = 0; date_idx < column_types.size(); date_idx++)
+    using PJ::CSV::ColumnType;
+    for (size_t index = 0; (index + 1) < column_types.size(); index++)
+    {
+      const auto date_idx = index;
+      const auto tie_idx = index + 1;
+      if((column_types[date_idx].type!= ColumnType::DATE_ONLY) ||
+         (column_types[time_idx].type!= ColumnType::TIME_ONLY))
+      {
+        continue; // early return is easier to read
+      }
+      // is_adjacent is implicitly true

Uh oh!

Add automatic detection and merging of separate date/time column (#1220) #1259

Are you sure you want to change the base?

Add automatic detection and merging of separate date/time column (#1220) #1259

Uh oh!

Conversation

lucabonamini commented Jan 7, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

lucabonamini commented Jan 7, 2026

Uh oh!

facontidavide left a comment

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

facontidavide commented Jan 10, 2026

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

lucabonamini commented Jan 10, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

lucabonamini commented Jan 7, 2026 •

edited

Loading