This turns issues like creating two shared ptrs from a single raw ptr
from popping up at runtime, instead making them a compile error.
Signed-off-by: flow <flowlnlnln@gmail.com>
While implementing the resource pack downloader in another branch, I
noticed that most of the code in the success callback was identical in
both cases, safe for a few minute differences in strings. So, this tries
to make it easier to share this piece of code.
However, it still leaves the possibility of extending the methods in
ResourceModel to accomodate for cases where this similarity may not
hold.
Signed-off-by: flow <flowlnlnln@gmail.com>
This allows us to check whether a search request is already on-going, in
which case we don't need to make another one (and shouldn't).
Signed-off-by: flow <flowlnlnln@gmail.com>
This prevents a crash in which the pack list gets updated in a search
request meanwhile a versions / extra info request is being processed.
Previously, this situation would cause the reference in the latter
callbacks to be invalidated by an internal relocation of the pack list.
Signed-off-by: flow <flowlnlnln@gmail.com>
This refactors the sorting methods to join every bit of it into a single
list, easing maintanance. It also removes the weird index contraint on
the list of methods by adding an index field to the DS that holds the
method.
Lastly, it puts the available methods on their respective API, so other
resources on the same API can re-use them later on.
Signed-off-by: flow <flowlnlnln@gmail.com>
This makes it so that we don't need a reference to the parent page in
the model. It will be useful once we change the page from a widget-based
one to a QML page.
It also makes tasks be created in the dialog instead of the page, so
that the dialog can also have the necessary information to mark versions
as selected / deselected easily. It also makes the task pointers into
smart pointers.
Signed-off-by: flow <flowlnlnln@gmail.com>
Puts them all inside the 'ResourceDownload' namespace, so that it's a
bit clearer from the outside that those belong to the same 'module'.
Signed-off-by: flow <flowlnlnln@gmail.com>
Firstly, this abstract away behavior in the mod download models that can
also be applied to other types of resources into a superclass, allowing
other resource types to be implemented without so much code duplication.
For that, this also generalizes the APIs used (currently, ModrinthAPI
and FlameAPI) to be able to make requests to other types of resources.
It also does a general cleanup of both of those. In particular, this
makes use of std::optional instead of invalid values for errors and,
well, optional values :p
This is a squash of some commits that were becoming too interlaced
together to be cleanly separated.
Signed-off-by: flow <flowlnlnln@gmail.com>
This allows us to pass to the creation instances their actual pack ID
and version ID, that in Flame's case, are only available before starting
to create an instance.
Signed-off-by: flow <flowlnlnln@gmail.com>
This would cause ODR violations when those headers were included in
other places that also included stuff like "Version.h" (note the
"meta/Version.h"), which can cause problems, especially in LTO.
Signed-off-by: flow <flowlnlnln@gmail.com>
Meaning we don't have to wait for the searches to finish in the
background to finally start the modpack download, when we have already
selected it -_-
Signed-off-by: flow <flowlnlnln@gmail.com>
This makes it similar to CF mods / modpacks. The mods cache is
maintained with the same name because it most likely has more data it
in, so this commit will affect existing caches as minimally as possible.
Signed-off-by: flow <flowlnlnln@gmail.com>
While working on pack updating, instance naming always gets in the way,
since we need both way of respecting the user's name choice, and a
standarized way of getting the original pack name / version.
This tries to circunvent such problems by abstracting away the naming
schema into it's own struct, holding both the original name / version,
and the user-defined name, so that everyone can be happy and world peace
can be achieved! (at least that's what i'd hope :c).
Signed-off-by: flow <flowlnlnln@gmail.com>
Since network requests are, for the most part, asynchronous, there's a
chance a request only comes through after the request sender has already
been deleted.
This adds a global (read static) hash table relating models for the mod
downloader to their status (true = alive, false = destroyed). It is a
bit of a hack, but I couldn't come up with a better way of doing this.
To reproduce the issue before this commit: scroll really quickly through
CF mods, to trigger network requests for their versions and description.
Then, in the middle of it close the mod downloader. Sometimes this will
create a crash.
Signed-off-by: flow <flowlnlnln@gmail.com>
It has a delay of 350ms from the last typed character to search, in
order to cache small changes while typing.
Signed-off-by: flow <flowlnlnln@gmail.com>
This allows us to define custom painting for list view items. In
particular, this is applied to the mod downloader, in order to allow
displaying both the mod name and mod description, and settings their
effects (like bold or underline) independent of each other.
Signed-off-by: flow <flowlnlnln@gmail.com>
