Using specific values for X axis (e.g. time) for plotting

[giuserive]

Hi,
I'm encountering a potential issue with the library when representing data in the time domain. Let me give an example: I have a dataset of size 1000x42. On the X-axis, there are 1000 values representing time. However, when displaying the data, it seems to consider the ordinal value rather than the actual value associated with that index.
For example, if I create a simple scatter plot in Excel, it looks like I'm getting the first plot, while I actually want the second. I'm also attaching the table with the values.

Am I doing something wrong?

I tried using the NX format—I correctly create the group and set the axes. It adds the labels and associates the corresponding value to each index, but it still seems like it's just a label rather than a real numerical axis.

Thanks in advice for help

[loichuder]

Hello @giuserive

Thanks for the report ! How do you store the time data in your HDF5 file ? It is using a special format like datetime or a string format ?

If the data is stored numerically, it should work fine (see the example below).

It could also come from the NX format: NeXus is quite picky and a single typo like axe instead of axes can break the visualization. You can share your file (or how you generate it) if you wish us to have a look at it.

import h5py

time1 = [12, 13, 14, 20, 21, 22, 23]
Y1 = [547.56, 595.36, 645.16, 696.96, 750.76, 806.56, 864.36]

with h5py.File("line.h5", "w") as h5file:
    line_grp = h5file.create_group("line")
    line_grp.attrs["NX_class"] = "NXdata"
    line_grp.create_dataset("time", data=time1)
    line_grp.create_dataset("Y1", data=Y1)
    line_grp.attrs["signal"] = "Y1"
    line_grp.attrs["axes"] = "time"

[giuserive]

Hello @loichuder,
thank you very much for your response. I’m attaching the file I started with and an example I generated using NX. The array containing the time values, which go on the X-axis, consists of integers (they are marked as such when inspected).

However, they are not in datetime format, so I’m not sure if they need to be converted. I’d like to ask if you could check their consistency.

This is the link: https://drive.google.com/file/d/1vXndCfMjWdoE1hbbFYcAr-KXmvowo0Hi/view?usp=sharing

[loichuder]

Thanks!

Your file looks fine to me and the generated plot as well. For example, if I look at the first slice along Y with respect to times (Z[:, 0] vs X), I get this in H5Web:

and this in matplotlib:

import matplotlib.pyplot as plt
from pathlib import Path
import h5py

with h5py.File(
    Path.home()
    / "CSES_01_EFD_1_L02_A1_056890_20190211_113815_20190211_121257_000_NX.h5",
    "r",
) as h5file:
    time = h5file["/A111_P/X"][()]
    z = h5file["/A111_P/Z"][:, 0]

plt.figure()
plt.plot(time, z)
plt.show()

---

Is this not what you expect ? Or do you get a different result ?

[giuserive]

Hi,
Thanks for the help!

It actually seems to work this way. I had some doubts because when I analyze the single dataset without grouping it in NX, it seems to behave like the example above, considering the ordinal value rather than the actual value of the node.

Maybe I should try with other files as well.

[loichuder]

when I analyze the single dataset without grouping it in NX, it seems to behave like the example above, considering the ordinal value rather than the actual value of the node.

Absolutely. When looking at a single dataset (no NX), H5web uses the indices (or ordinal value in your terms) as X axis values.

This is because there is no way for H5Web to guess what should be used as X axis values. This is not specific to H5Web, matplotlib does the same if no X axis is specified:

import matplotlib.pyplot as plt
from pathlib import Path
import h5py

with h5py.File(
    Path.home()
    / "CSES_01_EFD_1_L02_A1_056890_20190211_113815_20190211_121257_000_NX.h5",
    "r",
) as h5file:
    z = h5file["/A111_P/Z"][:, 0]

plt.figure()
plt.plot(z) # <--- No X
plt.show()

This is the job of the NX attributes to link axes to dataset values to produce the plots in my previous comment.

I'll close the discussion as resolved and rename it to better reflect our exchanges but don't hesitate to create a new one if you encounter another issue.