Trailing zero

[marrbnl]

Dear Developers

First of all, thanks very much for developing this very useful tool. I am currently using it for preparation of HEPData submission by converting txt files into yaml files. However, I encountered two related issues regarding trailing zeros.

1. A value of 4e-08 in txt file is converted into 4.0e-08 in yaml file
2. A value of 0.10 in text file is converted into 0.1 in yaml file

In both cases, the number of significant figures is changed, which is not what I want. Is there any way I can avoid this? Thanks a lot for your help.

[AndreasAlbert]

Could you share minimal code example for this? I suspect there might be two independent issues here:

1. If you load numbers into python floats, there is no real concept of "significant figures". float("0.10") and float("0.1") give the same result. The information you are trying to preserve is already lost here.
2. When hepdata_lib writes numbers into yaml, we round them to a configurable relative precision. By default, the number of significant digits is 5. You can change this behavior by e.g. setting my_variable.digits = 2. This will round all numbers in this Variable to the same relative number of significant digits (also counting digits before the decimal point!).

[marrbnl]

Thanks for coming back to me.

A minimal code example can be found here: https://www.dropbox.com/s/sb0d1f0k1cyzz1q/TestExample.zip?dl=0.

For fixed number of significant digits, it does not really serve my purpose since the same variable in different bins could have different numbers of significant digits. What I hope to achieve is to convert what's in the input the txt file exactly to the output yaml file.

[GraemeWatt]

I looked at the provided example. The numpy.loadtxt function uses a float data-type by default. You can instead load the data as strings to preserve the original precision: data = np.loadtxt(fname, dtype=str). I checked that the YAML file written by the example script with this modification then preserves the trailing zeros of the original text file.

[marrbnl]

I recompiled my ROOT with Python3, and can read out the txt file as str now. The str type is now propagated into the yaml file. Here is an example:
values:

• errors:
  • label: statistical errors
    symerror: '0.06'
  • label: systematic uncertainties
    symerror: '0.06'
    value: '0.65'

Is this fine? I mean there are quotation marks around the numbers. Thanks a lot for your help.

[GraemeWatt]

The quotation marks mean the value is stored as a string and not as a float. String values are supported in most fields of the data schema other than low and high bin values where you'd need to convert to a number (without quotation marks). The main HEPData code should convert the string values to numbers where needed, for example, to produce the online plots. You could upload an example to your Sandbox to check if it works.

[marrbnl]

I will check with the Sandbox. Thanks very much for all your help.

[clelange]

@marrbnl Could you please let us know if this fixes the problem for you? Then we could mark Graeme's answer as the solution.

[marrbnl]

After I converted string to numbers for low and high bin values, everything works fine in sandbox. Thanks for all the help.