Resolve #40 and demo for VL20

[yumengch]

Aim

Add bib and demo for VL20

Resolve issue #40

Potential issue

• Clean-up for function preprocess in align_col needed?
• 'frmt' in 'align_col' must be given carefully, not an issue for now.
• This used to be 0.03 now 0.030 based on prec though prec is not shown. I feel this is correct.

[patnr]

I feel like the combo unpack_uqs and align_col are too complicated. I will have a look at it after this, to see if I can simplify them.

Here are my changes to improve tabulation of results. https://github.com/nansencenter/DAPPER/compare/cd47cd0..e9c5535 Note that there are further improvements in the output of tests/test_data.py.

The main implementation changes are

• align_col no longer relies on tabulate. Instead it does the decimal-point alignment itself. This avoids roundabout pre/post-processing.
• All special formatting has been moved from align_col to unpack_uqs.
• All of the attrs of uqs are extracted by unpack_uqs.
• The above allowed to greatly simplify xp_process.py:xpSpace.print:make_cols

[yumengch]

Great work, Patrick!

Some comments:

• Assigning str type values to uq.val and uq.prec leads to error in uq.__str__. I'm not sure if it is necessary to safeguard uq.__str__ for the val and prec types. If this problem is limited within unpack_uq, I think we can simply ignore this.
• I did some tests for unpack_uqs:

  >>> import numpy as np
  >>> from dapper.tools.rounding import UncertainQtty as UQ
  >>> from dapper.stats import unpack_uqs, align_col
  # setup uq_list
  >>> vals = np.array([0.43085084, 0.9192679 , 0.87700933, 2., 0.1])
  >>> precs = np.array([0.08460475, 0.76110353, 0.001, 1.45934142, 1.35481388])
  >>> uq_list = [UQ(val, prec) for val, prec in zip(vals, precs)]
  # default unpack_uq
  >>> unpack_uqs(uq_list, decimals=None)
  {'val': ['0.43', '0.9', '0.877', '2', '0'], 'prec': ['0.08', '0.8', '0.001', '1', '1']}
  # specify decimals
  >>> unpack_uqs(uq_list, decimals=4)
  {'val': [0.4309, 0.9193, 0.877, 2.0, 0.1], 'prec': [0.0846, 0.7611, 0.001, 1.4593, 1.3548]}
  # check None array for TODO 3
  >>> col = unpack_uqs(np.array([None, None, None]), decimals=None)
  >>> align_col(['rand']+col['val'])
  ['rand', '␣␣␣␣', '␣␣␣␣', '␣␣␣␣']
  >>> align_col(['1s']+col['prec'])
  Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
  KeyError: 'prec'

  It is worth noting two special cases: -- 0.87700933 ± 0.001 and 2.0 ± 1.0 don't generate 0.8770±0.0010 and 2.0000±1.0000 when I specify decimals=4. This is again an issue of necessity to padding 0s for truncated values. -- Regarding your TODO 3: validate, the combination of unpack_uq and align_col in the xp_process should be fine, but may have error in tabulate_avrgs, but I'm not so sure if tabulate_avrgs needs to deal with such input.

[patnr]

Many thanks for the comments.

If this problem is limited within unpack_uq, I think we can simply ignore this.

It is contained within, yes, thanks to deepcopy. However, I changed this to vars(uq).copy() the most recent commit, which is maybe more explicit/safe solution.

don't generate 0.8770±0.0010 and2.0000±1.0000 when I specify decimals=4

Fixed in the latest commit. This fix was simple now that we don't have to worry about tabulate truncating zeros.

Regarding your TODO 3: validate

This is now gone, I believe. Not sure if a similar issue could exist. But it seems good enough for now.

[yumengch]

It looks great now. Thanks.