Independent Curves and Numerical Families

This guide covers GX-TXT Curve Digitalizer 0.9.8.10-dev-112, updated to patch P005. The control labels are given as they appear in the English interface.

This chapter shows how to organize curves before digitizing points: creating and editing independent series, labels and identifiers, visibility, numerical families, scrolling, deletion and saving. The five colored traces are part of the source image, not numerical data already digitized.

1. A Graph with Five Source Traces

The example uses a single canonical Image Dataset with the image.raster contract: a synthetic 1280 × 900 pixel PNG. The image contains five detector response traces, corresponding to five bias voltages.

Synthetic graph with five colored detector response traces for five bias voltages.
Figure 1. The five colored traces belong to the pixels of the original raster image, not to curves already digitized.

The figure was created for the tutorial, not acquired from an instrument. The raster curves have visual names and colors that are not automatically equivalent to the scientific objects of the Curves tab: these are created by the operator, separately.

2. The Scientific Axes Are Shared by the Curves of the Same Graph

In the Calibration panel, the physical axes Elapsed time, X=0–10 s, and Detector response, Y=−20–80 mV, were set, both linear. The calibration was acquired with the actual Set BL, Set BR, Set TL and Set TR buttons on the four corners of the plotted area.

Graph with four cyan calibration markers and Calibration ready.
Figure 2. The five series share a single projective transformation defined by the four calibration points.

In this test all five future curves will share the same geometrically calibrated transformation. There is no need to invent five independent homographies for five lines drawn on the same axes.

The association with the different bias values belongs to the family metadata, not to the four calibration corners.

3. The Curves Tab in an Empty Configuration

Opening Curves in the new configuration, there are the Curve family frame, the curve table, the New curve, Edit, Delete, Stroke width… and Show/Hide commands, as well as the controls for points and assisted proposals.

Empty Curves tab with Curve family frame, curve table and curve commands.
Figure 3. The image Dataset already contains colored traces, but no numerical series has been created yet.

The initial table contains no scientific curves. The source, however, already shows five colored traces: their presence on screen does not prove that five curves exist in the Workspace.

This distinction between the original raster and the Series objects is fundamental for interpreting the future Dataset Preview correctly as well.

4. New curve: Label and Optional Properties

Pressing New curve opens the native dialog with the fields:

Curve label, Family parameter value (optional) and Family unit (optional).

New curve dialog with Curve label, Family parameter value and Family unit fields.
Figure 4. The independent curve editor distinguishes the label from the optional fields of the numerical parameter.
New curve dialog with only the curve label filled in.
Figure 5. For independent curves, the family parameter value and its unit can be left empty.

When the configuration is not a numerical family, it is legitimate to create independent curves leaving the last two fields empty. For example, two traces that represent different measurements and are not parameterized by the same quantity must not be artificially forced into a family.

5. The Curve label Field Is Required

OK was pressed with the curve name left empty. The GUI showed the error:

Curve label is required.

Error message stating that the curve label is required.
Figure 6. The application rejects an empty label before creating a new curve.

No curve without a label was created. Entering Reference blue with the two family fields empty and pressing OK, the viewer created a curve with a stable identifier and zero points.

The display name must not be used as the scientific key of the Dataset: the model keeps a separate identifier for each curve.

6. Two Independent Curves in Generic Mode

A second curve, Reference red, was added without a numerical parameter. Two rows appeared in the table with an empty Family value and Points = 0.

Curves table with Reference blue and Reference red, empty family value and 0 points.
Figure 7. Two curves with distinct identifiers can exist without numerical family parameters.

The publication contract of the generic configuration is table.generic. The presence of the two curves does not yet imply a publication: no X/Y points have been acquired.

The colors the Viewer assigns to curve objects must not be confused with those of the source pixels, especially before digitization.

7. Show/Hide: Changing Visibility Does Not Mean Changing the Source

With the second curve selected, Show/Hide was pressed. The table added the text (hidden) to the curve label.

Curves table with the second curve marked hidden.
Figure 8. Show/Hide changes the visibility of the series without altering the pixels of the source PNG.
Curves table showing the hidden curve still listed.
Figure 9. The hidden curve stays in the table: no points or identifiers are deleted.

The viewer changed only the visible field of the selected curve. The scientific identifier, the number of points (still zero), the axes and the SHA-256 of the image stayed identical. The five colored lines contained in the PNG stayed visible because they are part of the original raster.

Pressing Show/Hide again, the curve became visible again. It was neither deleted nor duplicated.

8. When a Configuration Is a Numerical Family

The Curve family frame contains the This configuration is a numerical curve family checkbox, along with the Family quantity, Family unit and Semantic role fields.

For our example the following were used:

ControlValue
This configuration is a numerical curve familyEnabled
Family quantityGrid bias
Family unitV
Semantic rolefamily
Curve family frame with Grid bias, V and family set.
Figure 10. The numerical family control is used when the curves depend on a common physical quantity.

A numerical family makes sense when a common physical quantity parameterizes all the curves, such as grid voltage, temperature or concentration. It is not enough for the lines to be close together or to have different colors.

9. Enabling the Family Does Not Invent Missing Values

The family checkbox was enabled while there were still two curves with empty optional values. The GUI kept both rows, without automatically assigning bias numbers.

Curves table after enabling the family, with empty family values.
Figure 11. Enabling family mode does not automatically assign numerical values to existing curves.

The proposed output switched to the curve_family.table contract, visible in the Publish tab.

Publish tab showing the curve_family.table contract.
Figure 12. The contract intended for publishing the family is visible, but no canonical table has been created yet.

This step changes the intended publishing mode and semantics, but does not automatically fulfil the scientific requirements. Curve Digitalizer requires finite numerical values and consistent units before publishing a family.

10. Edit curve: Assigning the Family Values

Selecting the first curve and pressing Edit, Curve label = Bias -2 V, Family parameter value = -2 and Family unit = V were set.

Edit curve dialog with Bias -2 V, value -2 and unit V.
Figure 13. The curve identifier stays the same while the family parameter metadata are added.

The second curve was modified in the same way to Bias -1 V, value -1, unit V.

Curves table with Bias -2 V and Bias -1 V.
Figure 14. Two distinct series now share the Grid bias family quantity.

Comparing the Workspace before and after Edit confirmed that both curves keep the same identifiers. The names and family data changed, not the reference scientific curves or the source.

11. Completing the Five Curves

With New curve, three more series were created: Bias 0 V, Bias +1 V and Bias +2 V, each with the corresponding numerical value and the unit V.

New curve dialog creating a bias curve with its family value.
Figure 15. The new series receives a numerical parameter, not points extracted automatically from the image.
Curves table with five bias curves next to the five raster traces.
Figure 16. The five raster traces and the five scientific series remain different representations until points are digitized.

The result is a family of five Series objects, all distinct by ID and all still without points. The five traces visible in the image are available as a visual reference for the later digitization, but none has been converted automatically into X/Y data.

12. A Family Complete with Metadata

The final table of the Curves tab has the following rows:

Curve labelFamily valuePoints
Bias -2 V−2 V0
Bias -1 V−1 V0
Bias 0 V0 V0
Bias +1 V+1 V0
Bias +2 V+2 V0
Curves table with five bias curves from −2 V to +2 V and 0 points each.
Figure 17. All five series have family values, but contain no XY samples yet.

The viewer keeps these values inside the parameter metadata of each curve, in three distinct attributes: quantity name, value and unit. The group also has its own family definition with the enabled state and the semantic role.

A family value of 0 V is numerically valid: it must not be confused with a missing value.

13. Colored Trace and Scientific Curve Are Different Things

The source image shows five traces with different colors and a legend. The curve list shows five scientific objects with numerical bias values.

Graph with the five original raster traces and the legend.
Figure 18. The raster traces stay unchanged even when the properties and visibility of the series change.

Their equal number was built intentionally for the tutorial, but it does not prove that each object is already associated with the pixels of the right line. This correspondence will be built by digitizing points on the correct curve in the following chapters.

Even a visible curve with Points = 0 does not automatically produce a vector line of data: the drawing still belongs to the source PNG.

14. The curve_family.table Contract Is Not an Already Published Dataset

The Curve contract field in the Publish tab shows curve_family.table when the family is enabled. In non-family mode, the contract shown is table.generic.

The intended contract describes how the data would be published if numerical curves were ready, with valid family parameters and admitted calibrations.

15. Disabling the Family Keeps the Curves and Their Values

After the five series were created, This configuration is a numerical curve family was disabled.

Curve family frame with the numerical family checkbox disabled.
Figure 19. Generic mode uses the table.generic contract without deleting the curve parameters.

The mode became generic and the contract went back to table.generic.

Publish tab showing the table.generic contract.
Figure 20. The selected contract changes, but no Dataset is published and the source is not altered.

The five IDs, the five labels and the values −2, −1, 0, +1, +2 V stayed exactly as before. Disabling the family does not mean deleting the metadata of the individual curves, nor clearing the Series objects.

16. Re-enabling the Family Restores the Group Semantics

The family checkbox was then re-enabled in the same configuration, without recreating the curves.

Curves table after re-enabling the family, with the five bias curves unchanged.
Figure 21. Re-enabling the family does not recreate the series or change their identifiers.

This option lets you correct the classification of a configuration; it must not be used to arbitrarily give heterogeneous series a common numerical meaning.

17. Cancel in Edit curve Does Not Save the Draft

On the Bias 0 V curve, Edit curve was opened and a deliberately wrong name, value and unit were entered temporarily, without confirming them.

Edit curve dialog with temporary wrong values before Cancel.
Figure 22. Cancel discards the unconfirmed changes in the Edit curve dialog.

Pressing Cancel, the table and the Workspace keep the original values.

18. Changing Curve label Does Not Change the Curve ID

Opening Edit curve a second time, only the label was changed to Bias 0 V – checked, keeping the parameter 0 V.

Edit curve dialog renaming the curve to Bias 0 V - checked.
Figure 23. The series identifier and the numerical parameter are independent of the displayed label.
Curves table with the renamed Bias 0 V - checked curve.
Figure 24. Changing the label keeps the curve identifier.

In the run for this guide the label was then set back to Bias 0 V, in agreement with the legend of the image.

19. Long Lists Have Their Own Scrolling

To show the list scrolling, six empty temporary curves were created, named UI list stress 3 through UI list stress 8. They serve only to make the list long enough to scroll; they do not correspond to other traces in the image.

Curves table with eleven rows including the UI list stress curves.
Figure 25. Six empty temporary series make it possible to observe the list scrolling; they do not correspond to new raster traces.

The table thus has eleven rows: more than the nine that can be shown at once. These six temporary curves do not correspond to new traces in the image and are removed before the end of the chapter.

20. List Scrolling Is Independent of the Curves Page

The vertical control of the table was moved first to the beginning and then to the end of the list.

Curves table scrolled to the top of the list.
Figure 26. The curve list can scroll independently of the Curves page.
Curves table scrolled to the end of the list.
Figure 27. Scrolling the list does not change the position of the outer Curves page.

There are therefore two different scrollable areas: the curve list must not accidentally drag the whole control panel. Zooming the raster on the right is also a separate function.

21. Delete Requires Confirmation

Selecting the temporary curve UI list stress 8 and pressing Delete, the interface shows:

Delete curve ‘UI list stress 8’ and its 0 points?

Confirmation dialog asking to delete curve UI list stress 8 and its 0 points.
Figure 28. No keeps the selected curve; Yes deletes only the indicated temporary series.

Choosing No keeps the list of eleven curves. Repeating the operation and choosing Yes, only that temporary curve is deleted.

This command is stronger than Show/Hide and Cancel: it actually removes the selected scientific object. In our example the message shows 0 points, so no numerical data were deleted.

22. Cleaning Up the Temporary Curves and Checking the IDs

The other five curves created for scrolling are removed with the Delete command and the corresponding native confirmation.

Curves table back to the five bias curves.
Figure 29. The six temporary series were deleted without changing the identifiers of the five main curves.

The Workspace went back to containing exactly the five initial curves.

23. Back to the Five Labels of the Graph

The temporarily renamed curve was set back to Bias 0 V. All five labels in the table now match the values of the legend in the raster, with the unit V.

Curves table with the Bias 0 V label restored.
Figure 30. The label of the zero-bias curve was restored without changing its identifier or parameter.

This correspondence is an organizational choice for the example, not an automatic scientific measurement. To turn the lines into data you still need to select the correct curve and enter its points with the tools available in the Viewer.

24. Accepted Calibration and Numerically Empty Series

The already complete calibration of the same image was classified accepted with the Accept button of the Calibration tab.

Calibration measurements table with the measurement accepted.
Figure 31. Accept validates the geometric calibration, but does not create numerical XY points.

The result includes a group with one accepted calibration measurement and five parameterized series, but all with Points = 0. The accepted state of the calibration does not make the curves automatically ready for publishing.

This avoids a common confusion between approving the geometric transformation and accepting actually sampled curve data.

25. Publish: Observing the Contract Without Publishing

With the family enabled again, the Publish tab shows:

  • Curve contract = curve_family.table;
  • the Publication layout option with separate and combined;
  • one accepted calibration measurement among the eligible ones.
Publish tab with Publication layout separate and combined and one accepted calibration.
Figure 32. The Separate and Combined layouts are options available for publishing, which is not performed in this example.

The two layout options were only observed. Publish selected calibrations was not pressed, because there are no numerical points.

A later publication of numerical families requires finite values and consistent family units. This example does not publish anything and shows no resulting numerical Datasets.

26. Save, Close and Reopen: Checking Persistence

The final configuration was saved from the Configuration → Save configuration menu, then the Viewer was closed normally and reopened from the Viewers → Digitalizer catalog, in the same synthetic Session.

Reopened Viewer with the five bias curves and the family settings restored.
Figure 33. After reopening, the five curves, their identifiers and parameters, the numerical family and the accepted calibration remain.

The framework returned the same five identifiers, names and family values, the same Grid bias / V / family definition, the unchanged canonical source, the physical X/Y axes, the accepted calibration and the four corners.

The six temporary curves did not reappear. The Viewer stayed consistent with what had actually been saved, with no changes to the raster.

27. Meaning of the Data and Limits of the Test

This guide verifies the structure of the curves and the classification of the family, not the sampling or the publication of numerical data.

The scientific Series objects have distinct IDs and family metadata, but hold zero points. Consequently, only the source Image Dataset remains in the Session Dataset Catalog. There is not yet a canonical table of digitized data to compare with the Dataset Preview, nor a Plot Dataset API v4 result.

The visibility state concerns the presentation of the series, not the source pixels. The family value is a numerical property that identifies a described experimental condition, not a measure of uncertainty or a new observation of the Y response.

28. Essential Operating Procedure

For independent curves, create the objects with New curve, enter a label and leave the family fields empty when there is no common numerical quantity.

For a scientific family, enable This configuration is a numerical curve family and set Family quantity, Family unit and Semantic role. For each curve, enter a numerical value and the consistent unit through New curve or Edit curve.

Use Show/Hide to change only the presentation and Delete only when you really want to remove the series, checking the confirmation. Before closing, save the configuration and check that the IDs and family values have been kept.