Add DML support for samples

[EvanKirshenbaum]

While working on sensor support (#264), I'm realizing that it would be nice to be able to actually do something with the results in DML.

At the moment, you can say

s : take readings[([n], [interval/freq])]

and it writes to a CSV file, but you can't actually do anything with the readings. The operation just returns the number of readings taken.

What I'm thinking of doing is to say that what it returns is a Readings (subclassed by sensor, e.g., ESELogReadings) and that each of the attributes of this object is a T sample, which itself has attributes such as mean, min, max, count, variance, std dev, median, first, last. Other properties:

• A T sample can be implicitly converted to a T by taking the mean.
• Each of the attributes that would result in a divide by zero for an empty (or too small) sample returns a maybe value.
• You can add to a T sample by s : add(t).
• I'm probably going to have to split some sort of sampleable_param_type out from base_param_type, so I don't try to take the mean of a bunch of pads or bools.
  • I guess if I'm going to do that, I might as well split it further into quantity_param_type and component_param_type and numeric_param_type.
  • And I might as well get rid of the "param" notion.

    Migrated from internal repository. Originally created by @EvanKirshenbaum on May 31, 2023 at 4:20 PM PDT. Closed on Jun 22, 2023 at 2:58 PM PDT.

[EvanKirshenbaum]

This issue was referenced by the following commits before migration:

• 96090c071d994cbfbbc023c9af006566e00f2892 by @EvanKirshenbaum on Jun 02, 2023 at 12:26 PM PDT
• 2b07d5d29105b99db6f05d75cf8ea5d06d3f4de8 by @EvanKirshenbaum on Jun 19, 2023 at 3:56 PM PDT

[EvanKirshenbaum]

Okay, this all seems to be working. It's a bit clunkier than I would like on the Python end, because it turns out that I can't simply create a Sample[T] because there are actually three types involved:

• The item type (_ItemT), which is the type of the items added to the sample and the values returned by first, last, min_val and max_val.
• The difference type (_DiffT), returned by range, and std_dev.
  • This needs to be different from _ItemT for Sample[TemperaturePoint], where _DiffT is Temperature and for Sample[Timestamp], where _DiffT is Time.
• The continuous type (_ContT), returned by mean, median, harmonic_mean, and geometric_mean.
  • This needs to be different from _ItemT for Sample[int] where _ContT is float. So what I wound up with is Sample[_ItemT, _DiffT, _ContT], with subclasses IntSample, FloatSample, QuantitySample, TimestampSample, and TemperaturePointSample. To create samples, you can (and should) say, e.g.,

    s = Sample.for_type(int, (1, 5, 7))

    The overloads will tell you that what you get back will be an IntSample and that, therefore, the mean will be a float.

On the DML side, samples are referred to as, e.g., temperature sample, voltage sample, or float sample. To create them, you can say

s1 = an empty int sample;
s2 = a sample containing 3, 6, and 8.5;

(s2 will be inferred to be float sample.) For the containing expression, the and is optional, but if it's there and there are more than two elements, the Oxford comma is mandatory. (I couldn't figure out a way to get the grammar to not consider a, b and c to be a two element list where the last element is b and c, which requires to booleans.)

An int sample automatically converts to an int by taking the mean. Samples have the following attributes: count, first (first value), last (last value), min (minimum, min value, minimum value), max (maximum, max value, maximum value), range, mean (arithmetic mean), median, std dev (standard dev, std deviation, standard deviation), harmonic mean, and geometric mean.

You can add elements to a sample by

add(s, e)
s : add(e)

Migrated from internal repository. Originally created by @EvanKirshenbaum on Jun 22, 2023 at 2:58 PM PDT.