Be date spec comparator #226

treble37 · 2017-08-22T15:43:43Z

antonmi · 2017-10-03T15:01:26Z

Hi, @treble37 !
Sorry, I may have missed smth, but what is the status of this PR?
Is it ready for merging or you want to add smth?
Thanks!

treble37 · 2017-10-04T05:55:20Z

Hi @antonmi, just a few things...

I was hoping for feedback on the handling of the timezones wrt DateTime tests if someone is able... essentially what I did was backport a fair amount of the Elixir 1.5 functionality to make it compatible down to 1.3...
Shall I also add the corresponding ExUnit tests as well in this PR?
Based on the feedback, I might make some changes then, depending....

antonmi · 2017-10-04T05:57:23Z

Oh, sorry!
I will review soon

antonmi · 2017-10-04T11:56:43Z

Hi, @treble37 !
Sorry again for such late review!
There are a few comments:

Please namespace all the modules connected with this feature to separate module.
Let's say ESpec.DatesTimes. And place the code into dates_times directory.
So the structure will be:
lib/
dates_times/
comparator.ex (call the module ESpec.DatesTimes.Comparator, no need to separate Diff module)
date_time_protocol.ex (So module will be ESpec.DatesTimes.DateTimeProtocol )
types.ex (ESpec.DateTimes.Types )
impl/
date.ex
datetime.ex
naive_datetime.ex
time.ex
extensions/
calendar_extension.ex
datetime_extension.ex
integer_extension.ex
Please add specs for different cases. Place them to 'spec/dates_times' folder.
Please add README section which describes the cases

Also, could we think about improvements of be(:>, ...) and be(:<, ...) mathers? So they use the same comparison logic.

Thank you!

sascha-wolf · 2017-10-05T06:47:03Z

@antonmi, @treble37 are you okay with me chiming in on the changes and also providing a review? I don't want to force myself into this. 😉

antonmi · 2017-10-05T06:53:37Z

@zeeker you are very welcome!

treble37 · 2017-10-05T14:42:01Z

Hi @zeeker - your input would be welcome, I have a feeling there are things that can be improved, so please review away and don't be shy!

sascha-wolf · 2017-10-05T15:15:04Z

I will, lot's of meetings today! 😄

sascha-wolf

I'm asking myself if there is no native way of doing this. DateTime.diff/3 only works for seconds and below but isn't there something else?

Most of the stuff is of cosmetic nature, feel free to dismiss those.

sascha-wolf · 2017-10-06T07:10:17Z

lib/espec/comparator/diff.ex

+    end
+  end
+
+  defp do_diff(a, a, type),      do: zero(type)


The alignment of the do: seems to be off here.

Adjusted in accordance with #226 (comment) (If you use the do: syntax with functions and the line that makes up the function body is long, put the do: on a new line indented one level more than the previous line. )

sascha-wolf · 2017-10-06T07:10:17Z

lib/espec/assertions/be_close_to.ex

@@ -5,6 +5,12 @@ defmodule ESpec.Assertions.BeCloseTo do
  it do: expect(2).to be_close_to(1, 3)
  """
  use ESpec.Assertions.Interface
+  alias ESpec.Comparator.Diff


The community style guide suggests adding a blank line between directives.

Space added between use and alias

sascha-wolf · 2017-10-06T07:10:17Z

lib/espec/assertions/be_close_to.ex

+  alias ESpec.Comparator.Diff
+
+  defp match(subject, [value, {granularity, delta}] = data) do
+    result = abs(Diff.diff(subject, value, granularity)) <= delta


One could write that as follows but I'm not sure if that's a clearer approach; I'll mention it for completeness sake.

result = subject |> Diff.diff(value, granularity) |> abs() |> Kernel.<=(delta)

thanks, I like this and will use it.

sascha-wolf · 2017-10-06T07:10:17Z

lib/espec/assertions/be_close_to.ex

+
+  defp match(subject, [value, {granularity, delta}] = data) do
+    result = abs(Diff.diff(subject, value, granularity)) <= delta
+    {result, result}


Instead of return the result boolean twice you could return the result and the real delta. I'm aware that the original implementation didn't do that but I think it would be a great improvement!

Then you could use this in the success_message and/or error_message to inform the user about the real delta which might help in finding the issue.

Preliminary interface for providing a real computed delta with an error message added...

sascha-wolf · 2017-10-06T07:10:17Z

lib/espec/comparator/diff.ex

+  end
+  defp do_diff(a, b, :years) do
+    diff_years(a, b)
+  end


The community style guide suggests to only use single line dos if you have at maximum one multi line do. I would suggest changing this two to single line dos.

Adjusted in accordance with #226 (comment) (If you use the do: syntax with functions and the line that makes up the function body is long, put the do: on a new line indented one level more than the previous line. )

sascha-wolf · 2017-10-06T07:10:17Z

lib/espec/comparator/diff.ex

+    if a > b do
+      diff_years(end_date, start_date, 0)
+    else
+      diff_years(start_date, end_date, 0) * -1


Maybe also call this do_diff_years to match the approach of the diff_months implementation?

Will change, thanks!

sascha-wolf · 2017-10-06T07:10:17Z

lib/espec/comparator/diff.ex

+    diff_years(a, b)
+  end
+  defp do_diff(_, _, granularity) when not granularity in @units,
+    do: {:error, {:invalid_granularity, granularity}}


It also suggests to put all single line dos on a new line when you do it once.

Adjusted in accordance with #226 (comment) (If you use the do: syntax with functions and the line that makes up the function body is long, put the do: on a new line indented one level more than the previous line. )

sascha-wolf · 2017-10-06T07:10:17Z

lib/espec/comparator/diff.ex

+  defp do_diff(_, _, granularity) when not granularity in @units,
+    do: {:error, {:invalid_granularity, granularity}}
+
+  defp diff_years(a, a), do: 0


Is there actually a case where this will happen? Wouldn't this have been caught on line 23? Same for diff_months.

great catch Zeeker, will change

sascha-wolf · 2017-10-06T07:10:17Z

lib/espec/comparator/diff.ex

+          12
+      end
+    (year_diff*12)+month_diff
+  end


I'm not sure, but wouldn't it be possible to reuse the diff_years function to reduce complexity in do_diff_months?

@zeeker - If I'm understanding you correctly, the current implementation of diff_years that I think you're talking about (listed below) assumes a and b are date(times) converted to gregorian microseconds rather than integer years.....

defp diff_years(a, b) do {start_date, _} = :calendar.gregorian_seconds_to_datetime(div(a, 1_000*1_000)) {end_date, _} = :calendar.gregorian_seconds_to_datetime(div(b, 1_000*1_000)) if a > b do do_diff_years(end_date, start_date, 0) else do_diff_years(start_date, end_date, 0) * -1 end end

antonmi · 2017-10-06T08:54:03Z

@zeeker , thank you very much for detailed comments!

treble37 · 2017-11-05T00:44:59Z

Ok @antonmi (and @zeeker), this PR is getting close to done (other than some minor documentation cleanup), but I would like some feedback regarding your comment:

Also, could we think about improvements of be(:>, ...) and be(:<, ...) mathers? So they use the same comparison logic.

Commit 9b5d91d is my attempt to do this. What do you think of this API? If you're fine with it, then I will add the necessary tests to implement.

antonmi · 2017-11-05T16:35:35Z

Hi, @treble37 !

Thanks for your awesome work!
About be assertion:

What I meant was use the Comparator.diff when dates are passed to the be assertion:

Like this:

  defp match(%Date{} = subject, [op, %Date{} = val]), do: match_date_times(subject, [op, val])
  defp match(%Time{} = subject, [op, %Time{} = val]), do: match_date_times(subject, [op, val])
  defp match(%DateTime{} = subject, [op, %DateTime{} = val]), do: match_date_times(subject, [op, val])
  defp match(%NaiveDateTime{} = subject, [op, %NaiveDateTime{} = val]), do: match_date_times(subject, [op, val])

  defp match_date_times(subject, [op, val]) do
    delta = subject |> Comparator.diff(val, :seconds)

    result = apply(Kernel, op, [delta, 0])
    {result, result}
  end

But your API implementation is also very cool!
So, my suggestion: Let's keep both!

sascha-wolf · 2017-11-06T18:26:15Z

Hey @treble37,

I can only join antomni in saying thank you for your work and patience, and I'm glad I could be of some help.

I just have one final suggestion. In addition to the current be Implementation, you could add a case for a single keyword list, similar to how have can now receive values.

This would allow for the following syntax:

expect date |> to(be :>, other_date, years: 1)

I'll link in a second, I'm on mobile right now.

sascha-wolf · 2017-11-06T18:28:17Z

See commit ae7a4ca for details of what I mean.

treble37 · 2017-11-07T05:40:35Z

ok @zeeker, thank you for your review and suggestion(s) as always; regarding
your comment #226 (comment) this commit (ec3986d) enables it

I will work on adding the equivalent ExUnit tests and doing a final cleanup in the next 2 weeks or so then...

For issue antonmi#111 "Add Calendar expectation"

This reverts commit e694572. Import Timex logic to compare Date types and use a different API for Date type comparison

Change ESpec.Protocol to ESpec.DateTimeProtocol

Using the to_gregorian_microseconds interface - This is a bit misleading since we're really just converting the time to microseconds, it has nothing to do with gregorian

Add tests to be able to compare microsecond granularity for DateTime and NaiveDateTime

Avoid warnings such as: "undefined protocol function to_gregorian_seconds/1 (for protocol ESpec.DateTimeProtocol) lib/espec/datetime/datetime.ex:1"

Tests passing with Elixir 1.5.1 only

Tests passing for Elixir 1.4.5 (and 1.5.1) This commit is to backport Elixir 1.5.1 functionality via the Calendar.ISO.Extension module

This commit is to backport Elixir 1.5.1 functionality via the DateTime.Extension and Integer.Extension modules

Let's not rely on the version of Elixir to decide what method to call. Instead, we'll use one common method for all. Once ESpec migrates to supporting >= Elixir 1.5.x, then we can probably remove some of this supporting code.

Move this functionality into its own module, Espec.DatesTimesComparator

Help the user hunt down errors by providing the actual computed delta

Move file into dates_times folder for organizational clarity

Use the Comparator.diff when dates are passed to the be assertion. This lets us have the same comparison logic as the be_close_to assertion(s).

Enable testing syntax such as `expect date |> to(be :>, other_date, years: 1)`. Also, update tests to accomodate this new syntax.

Move one of the context descriptions in the describe text to avoid unneccessary block nesting

The dates_times files deal with assertions for Elixir's Date/DateTime/NaiveDateTime/Time types, so move it here.

Remove redundant variables

Also, fix typo in error message

Since we ported over Types from the Timex hex package, let's remove the unused types

treble37 · 2017-11-29T06:51:39Z

ok @antonmi and @zeeker, this PR is now updated with the functionality and I addressed yours and Zeeker's comments (which I put in the checklist up top #226 (comment)), so I guess it's ready to merge unless you have any other comments?

I can bump the version to ~~1.4.7~~ 1.5.0 and edit the CHANGELOG if you're ok with everything.

sascha-wolf · 2017-11-29T07:01:17Z

Awesome work @treble37! Semantic versioning would suggest to bump the version to 1.5.0 since you introduce a backwards compatible feature. Those are just my two cents though. 🙂

antonmi · 2017-11-29T09:26:15Z

Absolutely awesome!
Thank you @treble37 !
I will review and merge soon!
Thank you @zeeker for comments!

Update changelong in README Remove some minor warning messages as tests are run

treble37 · 2017-11-29T17:30:23Z

Of course you are right @zeeker - version bumped to 1.5.0 (maybe I shouldn't have had that last 🍷 😉 )

antonmi · 2017-11-30T09:01:40Z

ESpec 1.5.0 is live!
Thanks again to everyone!

treble37 mentioned this pull request Aug 22, 2017

Add Calendar expectation #111

Closed

sascha-wolf reviewed Oct 6, 2017

View reviewed changes

treble37 force-pushed the be-date-spec-comparator branch from f0196d5 to 04c7ef7 Compare October 8, 2017 22:50

treble37 force-pushed the be-date-spec-comparator branch from 9b2bbed to 5e4791b Compare October 22, 2017 23:56

treble37 force-pushed the be-date-spec-comparator branch from 9b5d91d to 44e885e Compare November 6, 2017 16:15

treble37 added 11 commits November 28, 2017 22:44

Add be_date() assertion

b819801

For issue antonmi#111 "Add Calendar expectation"

Revert "Add be_date() assertion"

19323fe

This reverts commit e694572. Import Timex logic to compare Date types and use a different API for Date type comparison

Add functionality and tests for NaiveDateTime

bf86c7e

Change ESpec.Protocol to ESpec.DateTimeProtocol

Add tests and functionality for Time

ef534a9

Using the to_gregorian_microseconds interface - This is a bit misleading since we're really just converting the time to microseconds, it has nothing to do with gregorian

Add tests and functionality for DateTime

414fd03

Add tests to be able to compare microsecond granularity for DateTime and NaiveDateTime

Remove unused protocol delegations

0311b0c

Avoid warnings such as: "undefined protocol function to_gregorian_seconds/1 (for protocol ESpec.DateTimeProtocol) lib/espec/datetime/datetime.ex:1"

Add DateTime time zone tests

e818120

Tests passing with Elixir 1.5.1 only

Add Calendar.ISO.Extension for Elixir 1.4.5

1a085f9

Tests passing for Elixir 1.4.5 (and 1.5.1) This commit is to backport Elixir 1.5.1 functionality via the Calendar.ISO.Extension module

Tests passing for Elixir 1.3.4 (and 1.4.5 and 1.5.1)

87f82d0

This commit is to backport Elixir 1.5.1 functionality via the DateTime.Extension and Integer.Extension modules

Update comments to extension modules

0e61f81

Move to_iso_days to Calendar.ISO.Extension

224a8c4

Let's not rely on the version of Elixir to decide what method to call. Instead, we'll use one common method for all. Once ESpec migrates to supporting >= Elixir 1.5.x, then we can probably remove some of this supporting code.

treble37 added 20 commits November 28, 2017 22:44

Separate functionality regarding Dates/Times

7e2f2b8

Move this functionality into its own module, Espec.DatesTimesComparator

Move DatesTimes related tests for be_close_to assertion

efbf056

Rewrite match to use pipe operator (squash)

1c7350a

Remove unused paths of diff_*(months, years) (squash)

f94d92b

Rename diff_years to do_diff_years to match implementation (squash)

5fdc556

Add Datetime comparison section to README

817a28d

Inform user about the real computed delta

53f9c9c

Help the user hunt down errors by providing the actual computed delta

Rename Delegator using a DatesTimes namespace

bc8271a

Move file into dates_times folder for organizational clarity

Add Error tests for DatesTimes BeCloseTo Assertion

21d4759

Add DateTime comparison logic for Assertions.Be

8c38769

Use the Comparator.diff to match be_close_to logic

e515ba0

Use the Comparator.diff when dates are passed to the be assertion. This lets us have the same comparison logic as the be_close_to assertion(s).

Allow Be comparison to take keyword list argument

395f148

Enable testing syntax such as `expect date |> to(be :>, other_date, years: 1)`. Also, update tests to accomodate this new syntax.

Remove double nested context block

5171897

Move one of the context descriptions in the describe text to avoid unneccessary block nesting

Move dates_times directory into spec/assertions

ee0d6a7

The dates_times files deal with assertions for Elixir's Date/DateTime/NaiveDateTime/Time types, so move it here.

Remove extraneous comment and update spec docs

5bb468d

Add ExUnit tests for BeCloseTo for DateTime(s)

3ebc5ae

Follow naming convention and rename be_close_to.exs

4f3a13c

Remove redundant variables

Add ExUnit tests for Be assertion for DateTime(s)

5da8999

Also, fix typo in error message

Update README with be assertion syntax

19c5d63

Remove unused Types

a99a448

Since we ported over Types from the Timex hex package, let's remove the unused types

treble37 force-pushed the be-date-spec-comparator branch from 9d14fca to a99a448 Compare November 29, 2017 06:48

treble37 added 2 commits November 29, 2017 09:25

Remove warnings caused by @impl

ba9f7bc

Bump version to 1.5.0

986d982

Update changelong in README Remove some minor warning messages as tests are run

antonmi merged commit ce848f5 into antonmi:master Nov 30, 2017

Be date spec comparator #226

Be date spec comparator #226

Conversation

treble37 commented Aug 22, 2017 • edited

antonmi commented Oct 3, 2017

treble37 commented Oct 4, 2017 • edited

antonmi commented Oct 4, 2017

antonmi commented Oct 4, 2017

sascha-wolf commented Oct 5, 2017 • edited

antonmi commented Oct 5, 2017

treble37 commented Oct 5, 2017

sascha-wolf commented Oct 5, 2017

sascha-wolf left a comment • edited

Choose a reason for hiding this comment

Choose a reason for hiding this comment

treble37 Oct 8, 2017 • edited

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

treble37 Oct 16, 2017 • edited

Choose a reason for hiding this comment

antonmi commented Oct 6, 2017

treble37 commented Nov 5, 2017

antonmi commented Nov 5, 2017

sascha-wolf commented Nov 6, 2017 • edited

sascha-wolf commented Nov 6, 2017 • edited

treble37 commented Nov 7, 2017 • edited

treble37 commented Nov 29, 2017 • edited

sascha-wolf commented Nov 29, 2017

antonmi commented Nov 29, 2017

treble37 commented Nov 29, 2017 • edited

antonmi commented Nov 30, 2017

treble37 commented Aug 22, 2017 •

edited

treble37 commented Oct 4, 2017 •

edited

sascha-wolf commented Oct 5, 2017 •

edited

sascha-wolf left a comment •

edited

treble37 Oct 8, 2017 •

edited

treble37 Oct 16, 2017 •

edited

sascha-wolf commented Nov 6, 2017 •

edited

sascha-wolf commented Nov 6, 2017 •

edited

treble37 commented Nov 7, 2017 •

edited

treble37 commented Nov 29, 2017 •

edited

treble37 commented Nov 29, 2017 •

edited