This implementation is based on the implementation in MantidPlot.
A C++-based executor has been added that can execute Python code
with the option of installing a trace function to update
a progress marker on a given editor window.

The old implementation attempted to chunk up the script into
the smallest executable blocks and mark each of these as when
they were executed. There were 2 issues:
  - the chunking of the scripts was error prone as it actually
    requires a good knowledge of how Python internally handles
    indentation etc. It was fairly easy to break and future
    internal updates to Python would have required unknown changes.
  - the progress reporting resolution was far poorer than MantidPlot
    as the actual line being executed was not reported but inferred
    by what block was being executed giving a false sense of progress
    through a script.

Switching to BytesIO and encoding the string means that cStringIO does
not need to be used which is not present in python 3.

Scripts containing an encoding declaration in the first two lines are
now able to be run without errors or encoding artefacts.

As the declaration cannot compiled properly is not encoded and causes
encoding artefacts if other parts containing unicode characters are
encoded, the solution has been to remove the declaration from the code
to be run if there would be a failure.

It was unclear (at least to me) what a particular line in
PythonCodeExecution.execute was doing. A comment was added so that no
one (like me) is tempted to remove it in the future.

The __future__ imports were not doing their job in scripts run in
Workbench because the relevant compiler flags were not passed to the
`compile` function when the scripts were executed.

Parse the given the given code for __future__ imports and pass the
relevant flags.

jedi crashes when launched using PyCharm's debugger on Python 2. Having
jedi as optional means we can still use the debugger whilst we're not
using Python 3.

Only add functions/builtins from simple api if they are title case and
don not begin with a '_'.

New CodeCompleter class contains the logic for populating QScintilla's
completion API. It's CodeCompleter.update_completion_api is called
whenever a script is successfully run.

re #25445

re #25140

if filename is set
Refs #24753

Update copyright headers in all files

Squashed commit of the following:

* First version of a script to manage copyright statements

re #23468

* Neatened up script, added comments

re #23488

* move script to tools directory

re #23488

* small script changes and a couple of manual file changes

re #23488

* Minor chnage to whitespace detection in regex

re #23488

* Add an excluded directory

re #23488

* remove a repeasted copyright statement in a file

re #23488

* Don't comsume the comment end if it is on the same line

re #23488

* fix error in new copright parsing

re #23488

* remove double copyrifght entry

re #23488

* Improve handling of old copyrights at the start of comments

re #23488

* remove empty comments re #23488

* exclude gsoapgenerated directories

re #23488

* Sort out greedy line matching re #23488

* improve empty comment removal

re #23488

* improve false positives

re #23488

* impressive speedup by limiting regex matching length

re #23488

* remove evil invisible non ascii character

Also upadte the copyright at the same time

re #23488

* resolve multiple copyrights in a single file

re #23488

* resolve an issue with new statement detection

re #23488

* another unprintable unicode character

re #23488

* pep updates and cmake the new copyright fit clang format

re #23488

* update already done new format headers

re #23488

* wrong type of bracket

re #23488

* Update class_maker and friends

re #23488

* Update all copyright statements

re #23488

* clang format re #23488

* flake8 warnings re #23488

* Flake8 warnings re #23488

* Exclude .cmake.in and rb.in files

re #23488

* replace missing line re #23488

* exclude .py.in files as they are flasely recognized as C++

re #23488

* another setp.py.in re #23488

* another .py.in correction re #23488

* Hopefully the last of the .py.in files re #23488

* resolve utf-8 encoding of python files and changed ABINS checksum

re #23488

* updates to unit tests that reference line numbers

re #23488

* remaining unit test files and other fixes

re #23488

Pete Peterson authored 6 years ago and WHITFIELDRE email committed 6 years ago

* Ignore syntax errors on startup

There was an uncaught exception in parsing the file on startup.

Refs #23555

* Invert the test that the code is executed

Refs #23555

._update_indent() was changed not to take arguments and IPython.core.usage.quick_guide was removed.

In test_execute_async_returns_failure_on_runtime_error_and_captures_expected_stack the comment was causing an IndentationError, changing it to a docstring fixes the problem. I don't think this is related to IPython but probably a python 3.6 problem.

Adds numpy and matplotlib to startup code.
Also silently executes a new interpreter as it starts so that the
completion api is populated.

Refs #21251

The IPython InputSplitter is assuming an interactive front end whereas
we have the whole code at once and do not have to anticipate users
possibly entering more. The new InputSplitter overrides
behaviour of IPython InputSplitter for these cases.
Refs #21251

Refs #21251

Refs #21251

Refs #21251

Refs #21251

Refs #21251

The marker should be placed at the end of the block not the start.
Refs #21251

It makes it easier to integrate with other code. This uncovered an
error in the code execution that has been fixed here. Previous changes
were not always executing the code as blocks and this caused
some tests to pass when they would fail in the block-based scenario.
We now only execute as blocks so the test now covers this.
Refs #21251

Refs #0

This effectively becomes the model for code execution within
a code editor.
Refs #21251

Refs #21251

Refs #21251

Refs #21251