#+PROPERTY: HEADER-ARGS+ :eval no-export :exports both
* Intro

An org document with code blocks to help test the proposed patches.

You need to load:
   #+begin_src elisp :results silent
   (load-file "my-async-tests.el")
   #+end_src



Emacs and org versions:
   #+begin_src elisp
   (mapcar (lambda (sb) (list sb (symbol-value sb)))
           '(emacs-version org-version))
   #+end_src

   #+RESULTS:
   | emacs-version | 30.0.50 |
   | org-version   | 9.7-pre |

Note that we've disabled eval on export: export doesn't know it needs
to wait for asynchronous results.

* A simple bash example
   :PROPERTIES:
    :header-args:bash: :execute-with my-shell-babel :nasync yes
   :END:

The package `my-async-tests.el' contains the function
`my-shell-babel-schedule' to evaluate shell script asynchronously.

The header-args properties above request asynchronous execution for
bash (:nasync yes), and, tells ob-core to use the prefix
`my-shell-babel' when looking for functions to evaluate a source
block. Thus, org will delegate execution to `my-shell-babel-schedule'.
We don't have `my-shell-babel-execute', so, in this case, :nasync must
be yes.

A simple execution:
    #+begin_src bash
      date
    #+end_src

    #+RESULTS:
    : Fri Feb 16 18:11:08 CET 2024

A tricky computation takes some time:
    #+begin_src bash
      sleep 5; date
    #+end_src

    #+RESULTS:
    : Fri Feb 16 17:58:23 CET 2024

An example of a failure:
    #+begin_src bash
      sleepdd 1; false
    #+end_src

    #+RESULTS:

* Python
   :PROPERTIES:
    :header-args:python:  :execute-with my-org-babel :nasync yes
    :header-args:python+: :session py-async
   :END:

Used =header-args= properties:
   - =:execute-with my-org-babel=:  look for functions with the prefix `my-org-babel' to execute
     blocks (for the asynchronous case use
     `my-org-babel-schedule', and, for the synchronous case
     `my-org-babel-execute'). These functions are defined in [[file:my-async-tests.el]].

   - =:nasync yes=: by default, execute asynchronously (use `my-org-babel-schedule').

   - =:session py-async= by default, use a session named "py-async".

** basic examples
*** async with a session
A very simple test:
    #+begin_src python
      2+3
    #+end_src

    #+RESULTS:
    : 5

Let's import the module time in our session.
    #+begin_src python :results silent
      import time
    #+end_src

    #+RESULTS:

(Yes, =:results silent= needs some work.)



A table that requires some time to compute:
    #+begin_src python
      start = time.time()
      time.sleep(1)
      end = time.time()
      ["%.1fs" % t for t in [start, end, end-start]]
    #+end_src

    #+RESULTS:
    | 1708103472.9s | 1708103473.9s | 1.0s |


An error (click on the error , <mouse-1>, to see the details):
    #+begin_src python
      2/0
    #+end_src

    #+RESULTS:


*** async with no session
   :PROPERTIES:
    :header-args:python+: :session none
   :END:

A very simple test:
    #+begin_src python
      2+3
    #+end_src

    #+RESULTS:
    : 5

Let's import the module time in our session.
    #+begin_src python :results silent
      import time
    #+end_src

    #+RESULTS:

(Yes, =:results silent= needs some work.)



A table that requires some time to compute:
    #+begin_src python
      start = time.time()
      time.sleep(1)
      end = time.time()
      ["%.1fs" % t for t in [start, end, end-start]]
    #+end_src

    #+RESULTS:
    | 1708083470.9s | 1708083471.9s | 1.0s |

Yes, it failed, as expected. "import time" was done in its own
temporary session.  The old result is preserved; the error is display
as an overlay. Click on it to get more info about the error.


Let's fix it, adding the import line:
    #+begin_src python
    import time
    start = time.time()
    time.sleep(1)
    end = time.time()
    ["%.1fs" % t for t in [start, end, end-start]]
    #+end_src

    #+RESULTS:
    | 1708102948.9s | 1708102949.9s | 1.0s |


An error (click on the error , <mouse-1>, to see the details):
    #+begin_src python
      2/0
    #+end_src

    #+RESULTS:



*** sync with a session
   :PROPERTIES:
    :header-args:python+: :session py-sync-session :nasync no
   :END:

A very simple test:
    #+begin_src python
      2+3
    #+end_src

    #+RESULTS:
    : 5

Let's import the module time in our session.
    #+begin_src python :results silent
      import time
    #+end_src

(Yes, =:results silent= needs some work.)



A table that requires some time to compute:
    #+begin_src python
      start = time.time()
      time.sleep(1)
      end = time.time()
      ["%.1fs" % t for t in [start, end, end-start]]
    #+end_src

    #+RESULTS:
    | 1708102997.5s | 1708102998.5s | 1.0s |



An error (click on the error , <mouse-1>, to see the details):
    #+begin_src python
      2/0
    #+end_src

    #+RESULTS:


*** sync with no session
   :PROPERTIES:
    :header-args:python+: :session none :nasync no
   :END:

A very simple test:
    #+begin_src python
      2+3
    #+end_src

    #+RESULTS:
    : 5

Let's import the module time in our session.
    #+begin_src python :results silent
      import time
    #+end_src

(Yes, =:results silent= needs some work.)



A table that requires some time to compute:
    #+begin_src python
      start = time.time()
      time.sleep(1)
      end = time.time()
      ["%.1fs" % t for t in [start, end, end-start]]
    #+end_src

    #+RESULTS:
    | 1708083470.9s | 1708083471.9s | 1.0s |

Yes, that fails (no session), displaying the details in a popup. Let's
fix it:
    #+begin_src python
    import time
    start = time.time()
    time.sleep(1)
    end = time.time()
    ["%.1fs" % t for t in [start, end, end-start]]
    #+end_src

    #+RESULTS:
    | 1708103039.0s | 1708103040.0s | 1.0s |



An error (click on the error , <mouse-1>, to see the details):
    #+begin_src python
      2/0
    #+end_src

    #+RESULTS:


** worg examples

Let's import matplotlib in our session.

    #+begin_src python
    import matplotlib
    import matplotlib.pyplot as plt
    #+end_src

    #+RESULTS:
    : None

A figure in a PDF, asynchronous case.
    #+begin_src python :results file link
    fig=plt.figure(figsize=(3,2))
    plt.plot([1,3,2])
    fig.tight_layout()

    fname = 'myfig-async.pdf'
    plt.savefig(fname)
    fname # return this to org-mode
    #+end_src

    #+RESULTS:
    [[file:myfig-async.pdf]]


A figure in a PDF, synchronous case.
    #+begin_src python :results file link :nasync no
    fig=plt.figure(figsize=(3,2))
    plt.plot([1,3,2])
    fig.tight_layout()

    fname = 'myfig-sync.pdf'
    plt.savefig(fname)
    fname # return this to org-mode
    #+end_src

    #+RESULTS:
    [[file:myfig-sync.pdf]]



A PNG figure, asynchronous case.
    #+begin_src python :results graphics file output :file boxplot.png
      fig=plt.figure(figsize=(3,2))
      plt.plot([1,3,2])
      fig.tight_layout()
      fig
    #+end_src

    #+RESULTS:
    [[file:boxplot.png]]

Same, but using the =:return= keyword.
    #+begin_src python :return "plt.gcf()" :results graphics file output :file boxplot.png
      fig=plt.figure(figsize=(3,2))
      plt.plot([1,3,2])
      fig.tight_layout()
    #+end_src

    #+RESULTS:
    [[file:boxplot.png]]

Same, asynchronous but without a session this time.
    #+begin_src python :return "plt.gcf()" :results graphics file output :file boxplot-no-sess-a-y.png :session none
      import matplotlib
      import matplotlib.pyplot as plt
      fig=plt.figure(figsize=(3,2))
      plt.plot([1,3,2])
      fig.tight_layout()
    #+end_src

    #+RESULTS:
    [[file:boxplot-no-sess-a-y.png]]


Lists are table,
    #+begin_src python
      [1,2,3]
    #+end_src

    #+RESULTS:
    | 1 | 2 | 3 |

unless requested otherwise.
    #+begin_src python :results verbatim
      [1,2,3]
    #+end_src

    #+RESULTS:
    : [1, 2, 3]


Dictionaries are tables too.
    #+begin_src python :results table
      {"a": 1, "b": 2}
    #+end_src

    #+RESULTS:
    | a | 1 |
    | b | 2 |


Let's try the example with Panda.
    #+begin_src python :results none
      import pandas as pd
      import numpy as np
    #+end_src

    #+RESULTS:
    : None

    #+begin_src python :results table
      pd.DataFrame(np.array([[1,2,3],[4,5,6]]),
                   columns=['a','b','c'])
    #+end_src

    #+RESULTS:
    |   | a | b | c |
    |---+---+---+---|
    | 0 | 1 | 2 | 3 |
    | 1 | 4 | 5 | 6 |

And the synchronous case?

    #+begin_src python :results table  :nasync no
      pd.DataFrame(np.array([[1,2,3],[4,5,6]]),
                   columns=['a','b','c'])
    #+end_src

    #+RESULTS:
    |   | a | b | c |
    |---+---+---+---|
    | 0 | 1 | 2 | 3 |
    | 1 | 4 | 5 | 6 |



Without session ?

    #+begin_src python :results table :session none
    pd.DataFrame(np.array([[1,2,3],[4,5,6]]),
                 columns=['a','b','c'])
    #+end_src

    #+RESULTS:
    |   | a | b | c |
    |---+---+---+---|
    | 0 | 1 | 2 | 3 |
    | 1 | 4 | 5 | 6 |

Right, we need to import the libraries (no session).

    #+begin_src python :results table :session none
    import pandas as pd
    import numpy as np
    pd.DataFrame(np.array([[1,2,3],[4,5,6]]),
                 columns=['a','b','c'])
    #+end_src

    #+RESULTS:
    |   | a | b | c |
    |---+---+---+---|
    | 0 | 1 | 2 | 3 |
    | 1 | 4 | 5 | 6 |


** inline examples

    A simple asynchronous inline src_python{3*2} {{{results(=6=)}}}.

    An other one containing a mistake src_python{2/0} {{{results(=6=)}}}
    (click on the error to see the details).


    Some very slow inline asynchronous computations that all run in
    the same session. You need to execute the 3 of them at once. Here
    is the first one src_python[:return "\"OK1\""]{import time;
    time.sleep(5)} {{{results(=OK1=)}}} and a second one
    src_python[:return "\"OK1 bis\""]{import time; time.sleep(5)}
    {{{results(=OK1 bis=)}}} and the third one src_python[:return
    "\"OK2\""]{import time; time.sleep(5)} {{{results(=OK2=)}}}.

    Yes, the previous paragraph is unreadable; it's on purpose, to
    check that ob-core can figure it out.

    Let's repeat, in a more readable way, and making the last one
    synchronous.

    Some very slow inline computations that all run in the same
    session. Here is the first asynchronous one
          src_python[:return"\"OK1\""]{import time; time.sleep(5)} {{{results(=None=)}}}
    and a second one, asynchronous too:
          src_python[:return "\"OK1 bis\""]{import time; time.sleep(5)} {{{results(=OK1 bis=)}}}
    and finally, a third one, synchronous this one:
          src_python[:nasync no :return "\"OK2\""]{import time; time.sleep(5)} {{{results(=OK2=)}}}.

    Note that, once the user executes the last synchronous block, the
    user is blocked until the synchronous execution can start
    (i.e. all previous asynchronous executions are done) and until
    it's done.  The display is updated though, to see the asynchronous
    progress.