Support significant digits in string representation of Measurements about pint HOT 9 CLOSED

I added formatting support to uncertainties. You can have a look at it in the GitHub version (https://github.com/lebigot/uncertainties/). The documentation is at https://github.com/lebigot/uncertainties/blob/master/doc/user_guide.rst#printing. I am ready to make adjustment to the formatting internals so that they can be used by Pint in a convenient way (I'm thinking of how parentheses are handled, here, as you may want to print like "(1.00±0.01)e+10 s", for instance, with no external parentheses, but also like "(1.00±0.01) s", with parentheses).

from pint.

It looks really good. How about if we try to agree on the formatting codes?

Pint uses:

• s for string
• r for repr
• l for latex
• p for pretty
• h for html (to be added soon)

Additionally, I use the modifier ~ to specify the short form of the units.

For what I can see, your u code is equivalent to what I propose to be s in this issue. s was not a good choice, I will happily change to u.

On the other hand, I like p more than C, because pretty print can be more than the plus/minus sign.

from pint.

Thanks. We can try to make the format specifications closer. :)

Yeah, s in your proposal mostly does what u does in uncertainties. s is indeed not perfect because it is used for string pre-formatting, not formatting ('{!s:10}'.format(1.23)).

Similarly, the usual notation for the representation is better as the usual '{!r}', I think.

I like p (pretty: ±) better than my C (character), so I will change C to P.

Now, I would like to raise the question of using uppercase letters for some options. I like the idea of interfering as little as possible with the usual float formatting, so I decided to use uppercase letters for the formatting options (L for LaTeX, P [now] for pretty-print and S for the shorthand notation). I kept the lowercase u for defining the number of significant digits on the uncertainty because it can appear before the float f/g/e format type, so that I find it natural to keep the first part of the format specification in lowercase, and have the options in uppercase. I take options to be anything that goes beyond what float formatting does (precision, total width, etc.).

What do you think about this uppercase question?

from pint.

Great!

In uppercase, there is also E, F and G. But I actually like L (uppercase) because l (lowercase) can be misread as I (uppercase i) depending on the font. So I am fine to change that.

I was playing a little bit with the representation of uncertainties. What do you think of this?

>>> '{}'.format(x)
0.200+/-0.010
>>> '{}'.format(x*1e7)
(2.00+/-0.10)e+06

>>> '{:.3u}'.format(x)
0.2000+/-0.0100
>>> '{:.3uP}'.format(x)
0.2000±0.0100
>>> '{:.3uL}'.format(x)
0.2000\pm0.0100

>>> '{:.3uP}'.format(x*1e7)
(0.2000±0.0100) 10⁶
>>> '{:.3uL}'.format(x*1e7)
\right(0.2000\pm0.0100\left) \times 10^{6}

>>> '{:(u)}'.format(x)
0.20(1)
>>> '{:(.3u)}'.format(x)
0.2000(0100)
>>> '{:(.3u)P}'.format(x)
0.2000(0100)
>>> '{:(.3u)L}'.format(x)
0.2000\left(0100\right)

>>> '{:(.3u)P}'.format(x*1e7)
0.2000(0100) 10⁶
>>> '{:(.3u)L}'.format(x*1e7)
0.2000\left(0100\right) \times 10^{6}

from pint.

The idea of using small exponent digits for the pretty-printing is interesting. I will probably include it in uncertainties. The only reservation I have now is that they might be printed too small in a terminal to be legible, but they do make sense. What do you think?

I think that the multiplication symbol should be added, in this case:

>>> '{:.3uP}'.format(x*1e7)
(0.2000±0.0100)×10⁶

no? This would be consistent with the current LaTeX mode too.

Note that '{:(.3u)}'.format(x) should print 0.2000(100) not 0.2000(0100): only the affected digits are written.

If I remember correctly "|eft(" and "\right)" is equivalent to "(" and ")", since what is inside the parenthese is small, so I believe that there is no need for \left and \right.

The '{:(u)}' is interesting, but it is also unusual (since it uses two symbol for a single formatting option), so I prefer something like for instance the current "S" (shorthand) option of uncertainties.

What do you think?

from pint.

Hi,

I like your version of pretty print with ×, and I agree with you about the S.

Regarding LaTeX, I use \left(and \right) because it forms a group. This is useful to prevent that is broken automatically into two lines when is near the end of one.

Ps.- You are of course right about the significant digits!

from pint.

I will start implementing this.

from pint.

I would suggest that you piggyback on uncertainties as much as possible: writing its formatting code was much more difficult and subtle than I initially thought… I will implement the new pretty-print (exponent digits, multiplication symbol) at some point, but I can't set a timeline, for now…

from pint.

Indeed it is. Most of the logic was implemented in Pint 0.2. I need to fix a few things and then dispatch to uncertainties the call. Do not worry about the timeline, I am really glad that we agreed on the specs.

from pint.