~/.aqlrc is just another aql script that can be loaded automatically
when the aql shell starts.

You can put standard aql commands inside the file.

Load the standard library files into Prompt.
The iteration order of the file list is unspecified[1] but I don't think
we have to deal with it anyhow. Even we create two files, say a.aql, and
b.aql and a function in b.aql uses a function from a.aql, the loading
order doesn't matter anyway.

Let's install the aql function files into DATADIR, which will probably
be /usr/share/algorithms-library/stdlib.

Let's start working on a standard library [1]. The idea is that we
bundle some aql files with predefined functions (and procedures) and we
load them when aql prompt starts.

So let's prepare a configuration file with a location for the scripts.
The location is currently only in local dev tree. We will later change
that to some /usr/share or /usr/lib location.

[1] #221

even though here the constant is not magical at all, it makes sense to have auto
here to let the array of escaped and actual special characters easily
extensible. Should that ever be needed.

Jan Trávníček authored 3 years ago and Jan Trávníček committed 3 years ago

breaking change!

In some cases, one wants to use system TClap (for instance, when one has
patched tclap installed)

Fixes: 7c0adf3e
Bug: #214

Long story short, tclap<=1.2.4 does not work with g++-11 compiler. We
sent a patch into TClap which was merged [1], however, new version was
not yet released so the distros have not included the fix in the
packaged versions of TClap.

This is becoming kind of a blocker for people who want to compile ALT.
I figured out that CMake can download external projects and if you do
things *magically right* [2] you can also convince the external projects
to be built before your project so you can depend on them.
So, CMake should now download TClap if you configure it to build aql2.
As tclap is header-only, compilation of aql2 modules should include the
directory where tclap is downloaded (and prefer it to the /usr/include
system path).
Unfortunately, tclap triggers a lot of warnings for us (hi,
-Wsuggest-override and -Werror) that we haven't seen before because
system headers do not trigger warnings. Fortunately, CMake can configure
includes to be SYSTEM includes (and use -isystem instead -I for g++
builds) which make things working again. However, this is just a really
big hack, AFAIK.

(I, personally, still think we should use some new, C++17-ready argument
parsing library. There is plenty of them.)

[1] https://sourceforge.net/p/tclap/code/ci/52a118c0edb20f64e8da88dc9d0bfee63f2232ac/
[2] https://crascit.com/2015/07/25/cmake-gtest/

Bug: #214
Bug: https://sourceforge.net/p/tclap/bugs/33/

the completion function gets int start and int end representing
boundaries of the completed text. The value is provided offset so there
should not be any need for negative values and I believe we may promote
that by internaly treating the values as unsigneds everywhere.

... because most developers were copy-pasting the preambule anyway and
did not pay any attention to change it.

Also, I think that git does better job in managing that.

Tomáš Pecka authored 4 years ago and Jan Trávníček committed 4 years ago

Although #pragma once is not a standard, compilers nowadays usually
support it. We don't have any tricky build system so we should be OK.

This also fixes many errors regarding using reserved identifiers (names
starting with double underscore).

Simplify versioning by CMake. Uses git version in the format
"<major>.<minor>.<patch>.r<commits-since-version-tag>.g<hash>(-dirty)?".

Remove CMakeLists.txt generator scripts from build phase. Replaced with
CMake's GLOB that can re-run every time the build is invoked. We can
also drop the Python dependency now.

Removed header installs which will be redesigned in future commits.