NPI-3429 - Fix mergesp3 utility

[ronaldmaj]

This PR is created to address the issue brought up here:
#28

Basically the mergesp3 utility in it's current form was not doing what it was supposed to.
Giving it a pair of any two sp3 files led to an error each time (a ValueError saying that the truth value of two series was ambiguous). This was coming up because in using the pd.concat function in gn_io.sp3.sp3merge there is a step where the equality of the df.attrs are checked. These are going to be different for sp3 files from different time periods.

Therefore, I've created a new variable that merges the attrs separately

merged_attrs = merge_attrs(sp3_dfs)

using an existing function merge_attrs, and then assigns the attrs of each df to an empty dict (thereby making them the same).

    for df in sp3_dfs:
        df.attrs = {}

One problem I encoutnered that took a while to debug was that even with no CLK files, I would keep ending up in the block of code that tries to incorporate CLK files into the dataframe, causing failure.
This is due to the mergesp3 utility producing an empty tuple rather than a None when no clkpaths are passed in.

I've manually set it to None now in that case.

[seballgeyer]

It would be interesting also to have some unit test on this one, just making sure that the output is what we expect.

[ronaldmaj]

It would be interesting also to have some unit test on this one, just making sure that the output is what we expect.

Yeah, good point, forgot about this. I'll put together something now - I need to get into the habit of writing unit tests when I do work on gnssanalysis 😄

[ronaldmaj]

It would be interesting also to have some unit test on this one, just making sure that the output is what we expect.

Okie doke, I've been able to write up a test for the sp3merge function. I did have to introduce a new py package unfortunately (pyfakefs) but that was the only way I saw to mimic a filesystem and therefore test the function itself.

Test seems to pass the pipeline ok:

[seballgeyer]

all good,
tested, pushed one correction wrt numpy v.2 compatibility

[treefern]

This will need updating again for Pandas 3. I'll try to replace it with format() calls at the point of writing the data out.

[ronaldmaj]

This will need updating again for Pandas 3. I'll try to replace it with format() calls at the point of writing the data out.

Yeah, there's a few changes that we'll need to make in order to make us Pandas 3 compliant. Running the sp3merge utility already has a whole bunch of warnings around this

[treefern]

Looks good to me.
I examined the padded date-stamp modifying line and couldn't think of a way it could go wrong.
It's nice to see some more unit tests coming in.