GitLab

Re #32125

Check for the new preprocess algorithm rather than the load algorithm
directly. Also we no longer have rename in the history because we are
only naming the algorithm at the last moment before putting it into the
ads with the correct name

Re #32125

Adds the missing property which exists in PreProcess to LoadAndProcess.

Re #32125

Return monitors when loading in the preprocess algorithm.
Re #32125

Switches to the new preprocess algorithm instead of loading directly.
Doesn't implement monitor handling currently.
Re #32125

Previously it was calling other algorithms directly which will not play
well with the history

Re #32125

Re #32125

Re #32125

Re #32125

Re #32125

Add outline algorithm and tests

Re #32125

The compiler potentially cannot make assumptions about the code and it
causes cppcheck to have a wobble since we "could" deref an invalid
iterator if we used it incorrectly

Fixes instances with missing assignment operator or copy constructor.
These are dangerous cases with dynamic memory as it's not obvious what
will happen, but the compiler will try anyway.

In many cases we can simply switch to unique_ptr or disable the copy
when it makes sense. But one or two cases required additional thought.

This fixes several memory leaks or sources of corruption if the fields
fall out of sync

These are various places where cppcheck now understands c++-17, or we
have improve our code

This should trigger a build whilst fixing a single warning

Signed-off-by: Jose Borreguero <borreguero@gmail.com>

Yuanpeng Zhang authored Sep 20, 2021 and Peterson, Peter committed Sep 21, 2021

Co-authored-by: Pete Peterson <peterfpeterson@gmail.com>

Rejects a duplicated kwargs and arg, as boost used to handle this before
we took over parsing manually using raw functions.
Now we will explicitly check for this, and throw a invalid_exception (we
couldn't find something for a TypeError though, so there is a potential
future improvement)

Workspace types were not accepted, this is particularly noticable in Fit
functions. Using the already-existing Workspace registry we can
trivially add a conversion that covers all of these types

We pass the correct type of size_t to raw_function but internally it
then uses it as an int, giving a compiler warning we can do nothing
about

Ports IFunctionAdaptor to use the same adaptor whilst preserving the
same input limitations.

Move the dynamic dispatcher into it's own class. This makes it trivial
to use this mechanism to dynamically sort user inputs and handle them in
various different ways

The compiler is smart enough to assign these types directly, this was a
workaround that was for the compiler not fully coercing the implicit
convertors for extractor types

IntArray is actually an array of longs, since we already have a long
type from Python we can preserve this until we try to set a scalar. In
the latter case we need to cast to an int for the underlying assignment
to work.

Boost will try to recurse down to the final level of a scalar value. In
our case we need to manually take over and dispatch for container types
using RTTI.

Uses recursive variant so supports lists of lists etc.