coffee/etl
1
0
Fork 0
mirror of https://github.com/ETLCPP/etl.git synced 2026-10-28 18:07:10 +08:00
Code Issues Packages Projects Releases Wiki Activity

Add docs and tutorial for etl::intrusive_avl_tree (#1608)

Browse Source 
* Add docs and tutorial for etl::intrusive_avl_tree

Documents the intrusive AVL tree added in PR #1425: a full API reference
page under Containers/Trees, and a tutorial demonstrating ordered access
to intrusively-linked objects with unpredictable lifetimes, including
move support and destructor-driven automatic removal from the tree.

Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>

* Add complexity shortcode and document its usage

Introduces a {{< complexity >}} Hugo shortcode (with light/dark styling
for constant/logarithmic/linear/linearithmic/polynomial/exponential
classes) for rendering Big-O complexity inline, applies it throughout
the intrusive_avl_tree API reference in place of raw "O(...)" text, and
documents the Description/Complexity structure and supported classes in
DOCUMENTATION.md.

Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>

* Fix issues found by @CodeRabbit

* Put complexity notes on their own lines in intrusive_avl_tree docs

Move the complexity shortcode from the inline "**Complexity:**" label
to its own line under a "**Complexity**" label. Notes that give several
alternatives (e.g. amortized vs. worst-case) now show each one on a
separate line. Update the guideline in DOCUMENTATION.md to match.

* Update complexity examples in DOCUMENTATION.md

Add the hard line break after the "**Complexity**" label in the
example, and rewrite the list of supported complexity classes in the
new two-line format, including an example with several alternatives.

* Add logarithmic complexity example to DOCUMENTATION.md

---------

Co-authored-by: Claude Sonnet 5 <noreply@anthropic.com>
Co-authored-by: John Wellbelove <jwellbelove@users.noreply.github.com>
This commit is contained in:
Sergei 2026-09-27 11:42:34 +03:00 committed by GitHub
parent d9a9488582
commit c313e6b98c
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
8 changed files with 894 additions and 0 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

32
DOCUMENTATION.md
View File

@ -120,6 +120,38 @@ The documentation is written in [Goldmark](https://github.com/teekennedy/goldmar
Hugo uses [Goldmark](https://github.com/teekennedy/goldmark-markdown) as its default Markdown processor for versions 0.60.0 and newer.
It is built into Hugo, providing high-performance and fully [CommonMark](https://commonmark.org/)-compliant rendering, along with support for [GitHub](https://docs.github.com/en/get-started/writing-on-github/getting-started-with-writing-and-formatting-on-github/basic-writing-and-formatting-syntax) Markdown.
### Documenting Algorithmic Complexity
When documenting the complexity of an operation, use the `complexity` shortcode instead of writing raw Big-O text.
Place it on its own `**Complexity**` lines, right after the `**Description**` line/paragraph, for example:
```markdown
**Description**
Returns the number of linked items.
**Complexity**
{{< complexity "constant" "1" >}}
```
The shortcode takes two arguments: a complexity class (used to color and label it) and the expression to show
inside `O(...)`. The supported classes are:
```markdown
**Complexity**
{{< complexity "constant" "1" >}}
**Complexity**
{{< complexity "linear" "N" >}}, where N is number of items in the tree, and assuming comparator is {{< complexity "constant" "1" >}}.
**Complexity**
{{< complexity "logarithmic" "log N" >}}, where N is number of items in the tree.
**Complexity**
{{< complexity "linearithmic" "N log N" >}}, where N is number of items in the tree.
**Complexity**
{{< complexity "polynomial" "N²" >}}, where N is number of items in the tree.
**Complexity**
{{< complexity "constant" "1" >}} when something, or
{{< complexity "exponential" "2ᴺ" >}} some alternative, where N is number of items in the tree.
```
### Starting a New Page
A template file is provided at `docs/page-template.md` to use as a starting point. Copy it to the

6
docs/containers/trees/_index.md Normal file
View File

@ -0,0 +1,6 @@
---
title: "Trees"
weight: 100
---
Tree like containers.

560
docs/containers/trees/intrusive-avl-tree.md Normal file
View File

@ -0,0 +1,560 @@
---
title: "intrusive_avl_tree"
weight: 1
---
{{< callout type="info">}}
Header: `intrusive_avl_tree.h`
Similar to: `std::set` / `std::map` (ordering only, no ownership)
{{< /callout >}}
An intrusive, self-balancing AVL tree.
```cpp
template <typename TValue, size_t ID_ = 0>
etl::intrusive_avl_tree
```
`TValue` is the type that contains the actual values. It must derive from
`etl::intrusive_avl_tree_base<ID_>::link_type`, exposed as the tree's own
`link_type`. Unlike `etl::intrusive_list`/`etl::intrusive_forward_list`,
which are templated directly on a link type and expect the node to derive
from it publicly, `intrusive_avl_tree` is templated only on `TValue` and an
`ID_`. Its `link_type` is a deliberately opaque wrapper around
`etl::tree_link<ID_>` that exposes almost nothing as public (or even
protected) API, by design, so that inheriting from it imposes as little
"intrusive"-ness as possible, avoids name conflicts, and hides implementation
details.
As with the other intrusive containers, an object may inherit multiple
`link_type`s with different `ID_` values to be a member of several trees (or
lists) at once:
```cpp
typedef etl::intrusive_avl_tree_base<0>::link_type FirstLink;
typedef etl::intrusive_avl_tree_base<1>::link_type SecondLink;
struct Item : public FirstLink, public SecondLink
{
int value;
};
etl::intrusive_avl_tree<Item, 0> tree0;
etl::intrusive_avl_tree<Item, 1> tree1;
```
See also:
[Intrusive Links]({{% relref "../../utilities/intrusive-links.md" %}})
[intrusive_list]({{% relref "../lists/intrusive-list.md" %}})
[intrusive_forward_list]({{% relref "../lists/intrusive-forward-list.md" %}})
There is no class-level comparator template parameter. Every ordering
operation instead takes its own comparator functor at the call site.
Lookup functions (`find`, `lower_bound`, `upper_bound`, `find_or_insert`)
take a unary comparator accepting a single `const value_type&` and returning
an `int`: `>0` if the search target is "greater" than the argument, `0` if
found, `<0` if "smaller". The range constructor takes a binary comparator
`int(const value_type& lhs, const value_type& rhs)`; `etl::compare<value_type>::cmp`
can be used directly if `value_type` defines `<`.
## Member types
```cpp
link_type (nested, inherited from etl::intrusive_avl_tree_base<ID_>)
value_type TValue
pointer value_type*
const_pointer const value_type*
reference value_type&
const_reference const value_type&
size_type size_t
rvalue_reference value_type&& // C++11
```
## The link type
`TValue` must derive from `link_type`. It has its own set of special member
functions that a node author relies on.
```cpp
link_type()
```
**Description**
Default constructor. Leaves the item unlinked.
---
```cpp
link_type(link_type&& other)
```
**Description**
C++11. Constructs a new item by moving `other` into `this`. After
construction, `this` replaces `other` in the same tree position that
`other` occupied, so no tree balancing is needed. `other` becomes unlinked.
**Complexity**
{{< complexity "constant" "1" >}}
---
```cpp
link_type& operator=(link_type&& other)
```
**Description**
C++11. Assigns `other` by moving it into `this`. Does nothing on
self-assignment. After assignment, `this` replaces `other` in the same tree
position that `other` occupied; `other` becomes unlinked. `this` might end
up in a different tree than it started in.
**Complexity**
{{< complexity "logarithmic" "log N" >}} if `this` is already linked to a tree (it has to be erased first, with rebalancing), or
{{< complexity "constant" "1" >}} if `this` is not linked, where `N` is number of items in the tree.
---
```cpp
link_type(const link_type&) = delete;
link_type& operator=(const link_type&) = delete;
```
**Description**
C++11. Copy construction and assignment are disabled.
---
```cpp
~link_type()
```
**Description**
Might rotate the tree as necessary to keep it balanced. The tree is not the
real owner of its nodes' memory - the node's actual owner may destroy it at
any time, with or without a prior explicit `erase`. If the link is still
linked to a tree when it's destroyed, the destructor unlinks it
automatically; otherwise the node's former parent and children would keep
dangling pointers to it, breaking the tree and leading to undefined
behaviour. This is what makes it safe to embed a tree node in an object
whose lifetime is managed elsewhere.
**Complexity**
{{< complexity "logarithmic" "log N" >}}, where `N` is number of items in the tree.
## Constructors
```cpp
etl::intrusive_avl_tree<TValue, ID_>()
```
**Description**
Default constructor. Constructs an empty tree.
---
```cpp
template <typename TIterator, typename TBinaryCompare>
etl::intrusive_avl_tree<TValue, ID_>(TIterator first, TIterator last, TBinaryCompare binary_comp)
```
**Description**
Constructs the tree from the range [`first`, `last`) of items, using
`binary_comp` to order them. All items in the range must be unlinked
initially; if any item is already linked to some other tree, throws
`etl::intrusive_avl_tree_value_is_already_linked`, and any items already
inserted are unlinked again - it's all or nothing.
If duplicates should be kept, `binary_comp` should return a non-zero result
even for "equal" items - `<0` prepends a duplicate, `>0` appends one.
Returning `0` for equal items keeps only the first of each run of duplicates
in the range; the rest are left unlinked. `etl::compare<value_type>::cmp`
can be used directly as `binary_comp`.
**Complexity**
{{< complexity "linearithmic" "N log N" >}}, where `N` is number of items in the range.
---
```cpp
intrusive_avl_tree(intrusive_avl_tree&&) = default;
```
**Description**
C++11. Move constructor.
**Complexity**
{{< complexity "constant" "1" >}}
---
```cpp
intrusive_avl_tree(const intrusive_avl_tree&) = delete;
```
**Description**
C++11. Copy construction is disabled.
---
```cpp
~intrusive_avl_tree()
```
**Description**
Every remaining node is unlinked; none of the value objects are destroyed,
since the tree never owns their memory.
**Complexity**
{{< complexity "linear" "N" >}}, where `N` is number of items in the tree.
## Assignment
```cpp
intrusive_avl_tree& operator=(intrusive_avl_tree&&) = default;
```
**Description**
C++11. Move assignment. Does nothing on self-assignment. All of `this`
tree's former items have to be unlinked.
**Complexity**
{{< complexity "linear" "N" >}}, where `N` is the size of `this` tree before assignment.
---
```cpp
intrusive_avl_tree& operator=(const intrusive_avl_tree&) = delete;
```
**Description**
C++11. Copy assignment is disabled.
## Capacity
```cpp
bool empty() const
```
**Description**
Returns `true` if the tree has no linked items.
**Complexity**
{{< complexity "constant" "1" >}}
---
```cpp
size_t size() const
```
**Description**
Returns the number of linked items.
**Complexity**
{{< complexity "constant" "1" >}}
## Iterators
Only bidirectional `iterator`/`const_iterator` are provided; there is no
`reverse_iterator` (use `visit_in_order` with `is_reverse = true` instead).
An iterator may be in one of three logical states: it references a real
node (`has_value() == true`), it is the terminal sentinel (`it == end()`),
or it is "valueless" (`!has_value() && it != end()` - e.g. a
default-constructed iterator, or the result of `get_parent()`/`get_child()`
when there is no such node).
```cpp
iterator begin()
const_iterator begin() const
const_iterator cbegin() const
```
**Description**
Returns an iterator to the smallest item. Returns `end()` if the tree is
empty.
**Complexity**
{{< complexity "logarithmic" "log N" >}}, where `N` is number of items in the tree.
---
```cpp
iterator end()
const_iterator end() const
const_iterator cend() const
```
**Description**
Returns an iterator to the terminal sentinel.
**Complexity**
{{< complexity "constant" "1" >}}
---
```cpp
iterator min()
const_iterator min() const
```
**Description**
Equivalent to `begin()`.
**Complexity**
{{< complexity "logarithmic" "log N" >}}, where `N` is number of items in the tree.
---
```cpp
iterator max()
const_iterator max() const
```
**Description**
Equivalent to `--end()`.
**Complexity**
{{< complexity "logarithmic" "log N" >}}, where `N` is number of items in the tree.
---
```cpp
iterator& operator++()
iterator operator++(int)
```
**Description**
Advances the iterator to the next ("greater") item.
**Complexity**
{{< complexity "constant" "1" >}} amortized,
{{< complexity "logarithmic" "log N" >}} worst-case, where `N` is number of items in the tree.
---
```cpp
iterator& operator--()
iterator operator--(int)
```
**Description**
Moves the iterator to the previous ("smaller") item.
**Complexity**
{{< complexity "constant" "1" >}} amortized,
{{< complexity "logarithmic" "log N" >}} worst-case, where `N` is number of items in the tree.
---
```cpp
reference operator*() const
pointer operator->() const
pointer get() const
```
**Description**
Dereferences an iterator that has a value. `operator*`/`operator->` throw
`etl::intrusive_avl_tree_iterator_exception` if `has_value()` is `false`.
`get()` returns `nullptr` instead of throwing.
---
```cpp
bool has_value() const
explicit operator bool() const
```
**Description**
Returns `true` if the iterator references a real node.
## Advanced traversal
These are normally only needed for advanced traversal.
```cpp
iterator get_root()
const_iterator get_root() const
```
**Description**
Returns an iterator to the root node. Valueless (`has_value() == false`) if
the tree is empty.
**Complexity**
{{< complexity "constant" "1" >}}
---
```cpp
int_fast8_t get_balance_factor() const
```
**Description**
Returns the node's AVL balance factor: `-1`, `0` or `+1`.
**Complexity**
{{< complexity "constant" "1" >}}
---
```cpp
iterator get_parent() const
```
**Description**
Returns an iterator to the node's parent. Valueless if there is no parent.
**Complexity**
{{< complexity "constant" "1" >}}
---
```cpp
iterator get_child(bool is_right) const
```
**Description**
Returns an iterator to the node's left (`is_right == false`) or right
(`is_right == true`) child. Valueless if there is no such child.
**Complexity**
{{< complexity "constant" "1" >}}
## Lookup
```cpp
template <typename TCompare>
iterator find(TCompare comp)
template <typename TCompare>
const_iterator find(TCompare comp) const
```
**Description**
Finds an item using the unary comparator `comp`. Returns `end()` if there
is no matching item.
**Complexity**
{{< complexity "logarithmic" "log N" >}}, where `N` is number of items in the tree,
and assuming `comp` is {{< complexity "constant" "1" >}}.
---
```cpp
template <typename TCompare>
iterator lower_bound(TCompare comp)
template <typename TCompare>
const_iterator lower_bound(TCompare comp) const
```
**Description**
Returns an iterator to the first item that compares as "not less" than the
target of `comp`. Returns `end()` if no such item exists.
**Complexity**
{{< complexity "logarithmic" "log N" >}}, where `N` is number of items in the tree,
and assuming `comp` is {{< complexity "constant" "1" >}}.
---
```cpp
template <typename TCompare>
iterator upper_bound(TCompare comp)
template <typename TCompare>
const_iterator upper_bound(TCompare comp) const
```
**Description**
Returns an iterator to the first item that compares as "greater" than the
target of `comp`. Returns `end()` if no such item exists.
**Complexity**
{{< complexity "logarithmic" "log N" >}}, where `N` is number of items in the tree,
and assuming `comp` is {{< complexity "constant" "1" >}}.
## Modifiers
```cpp
template <typename TCompare, typename TFactory>
etl::pair<iterator, bool> find_or_insert(TCompare comp, TFactory factory)
```
**Description**
Finds an existing item using the unary comparator `comp`; if not found,
calls `factory()` (no arguments) and inserts the item it returns at the
position where the search stopped. This operation never invalidates
existing iterators, though depending on where the new item was linked,
existing iterators may skip over it.
`factory` must return the address of the new value (castable to
`link_type*`). If it returns `nullptr`, the tree is left unmodified and the
result iterator is valueless. If the returned value is already linked to
any tree, throws `etl::intrusive_avl_tree_value_is_already_linked`.
Returns a pair of: an iterator to the found or inserted item (never `end()`,
but may be valueless), and a `bool` that is `true` if the tree was modified.
If duplicates are wanted, `comp` should return a non-zero result even for
"equal" items - `+1` appends after existing duplicates, `-1` prepends
before them.
**Complexity**
{{< complexity "logarithmic" "log N" >}}, where `N` is number of items in the tree,
and assuming `comp` and `factory` are {{< complexity "constant" "1" >}}.
---
```cpp
iterator erase(iterator position)
iterator erase(const_iterator position)
```
**Description**
Erases the item at `position`, including tree rebalancing. `position` must
reference a real item (asserts/throws
`etl::intrusive_avl_tree_iterator_exception` otherwise) and must originate
from this tree instance. Invalidates any existing iterator to the erased
item, but no others. Returns an iterator to the next node.
Use `clear()` instead if every item needs to be erased - no rebalancing is
involved.
**Complexity**
{{< complexity "logarithmic" "log N" >}}, where `N` is number of items in the tree.
---
```cpp
void clear()
```
**Description**
Unlinks every item, leaving the tree empty. Invalidates all existing
iterators. This is cheaper than calling `erase` on every item, which would
cost {{< complexity "linearithmic" "N log N" >}} once intermediate
rebalancing is taken into account.
**Complexity**
{{< complexity "linear" "N" >}}, where `N` is number of items in the tree.
---
```cpp
void swap(intrusive_avl_tree& other)
friend void swap(intrusive_avl_tree& lhs, intrusive_avl_tree& rhs)
```
**Description**
Swaps the contents of two trees. Does nothing on self-swap.
**Complexity**
{{< complexity "constant" "1" >}}
## Traversal
Each of these visits every item exactly once, without recursion, and has a
`const` overload. The visitor must not modify the tree during visitation.
```cpp
template <typename Visitor>
void visit_in_order(bool is_reverse, Visitor visitor)
```
**Description**
Visits every item in sorted order. `is_reverse` selects ascending (`false`)
or descending (`true`) order.
**Complexity**
{{< complexity "linear" "N" >}}, where `N` is number of items in the tree.
---
```cpp
template <typename Visitor>
void visit_pre_order(bool is_reverse, Visitor visitor)
```
**Description**
Visits every item in pre-order (parent before children).
**Complexity**
{{< complexity "linear" "N" >}}, where `N` is number of items in the tree.
---
```cpp
template <typename Visitor>
void visit_post_order(bool is_reverse, Visitor visitor)
```
**Description**
Visits every item in post-order (children before parent).
**Complexity**
{{< complexity "linear" "N" >}}, where `N` is number of items in the tree.
## Exceptions
```cpp
intrusive_avl_tree_exception // base exception
intrusive_avl_tree_iterator_exception // invalid iterator use
intrusive_avl_tree_value_is_already_linked // item already linked to a tree
```

5
docs/tutorials/_index.md
View File

@ -13,6 +13,11 @@ A quick introduction to containers.
---
[Intrusive AVL Tree](./intrusive-avl-tree-tutorial)
Ordering objects that already exist, without a second container.
---
[observer](./observer-tutorial)
A templated set of classes to allow easier implementation of the observer pattern.

210
docs/tutorials/intrusive-avl-tree-tutorial.md Normal file
View File

@ -0,0 +1,210 @@
---
title: "Intrusive AVL Tree"
---
A tutorial on how `etl::intrusive_avl_tree` can give ordered,
{{< complexity "logarithmic" "log N" >}} access to objects that already
exist elsewhere, without a second container to track them.
## Overview
Imagine a set of driver/sensor tasks, each owned by its own subsystem. They
are created and destroyed at unpredictable times as modules attach and
detach, so there is no compile-time upper bound on how many exist at once.
A scheduler needs to visit the tasks that are currently due, in ascending
deadline order, on every tick.
A normal fixed-capacity ordered container (`etl::set`/`etl::map`) would need
a second, separately-sized place to store pointers to these tasks. Instead,
embedding an `etl::intrusive_avl_tree` link directly into each task object
gives the scheduler ordered, {{< complexity "logarithmic" "log N" >}}
access to exactly the tasks that exist right now - nothing more is
allocated, and nothing needs to be kept in sync by hand. That's the payoff of "intrusive": the tree links the
objects that already exist, rather than storing copies of them.
The example below also demonstrates two things that fall out of the design
for free:
- **Move support.** When the `std::vector` that owns the tasks reallocates,
every task is moved into new storage - and each move takes over its
task's exact tree position, with no rebalancing and no code of its own.
- **Safe removal without an explicit `erase()` call.** Removing a task's
owning object removes it from the schedule too, with no dangling entry
left behind - whether that's because the object is destroyed while still
linked (`link_type`'s destructor unlinks it automatically), or, as in
this example, because `std::vector::erase` move-assigns another task over
it (`link_type`'s move assignment unlinks `this` before taking over the
moved-from tree position).
## Example
```cpp
#include "etl/intrusive_avl_tree.h"
#include <cstdint>
#include <iostream>
#include <vector>
#include <algorithm>
// A driver/sensor task that already exists as a long-lived object owned by
// its own subsystem. Inheriting from the tree's link_type is all it takes
// to make it schedulable - no separate fixed-capacity container is needed
// to track which tasks exist.
class SensorTask : public etl::intrusive_avl_tree_base<0>::link_type
{
public:
SensorTask(uint32_t id_, uint32_t deadline_, uint32_t period_)
: id(id_), deadline(deadline_), period(period_)
{
}
// Orders primarily by deadline; ties are broken by id so two tasks due
// on the same tick still occupy distinct tree positions.
struct Locate
{
uint32_t target_deadline;
uint32_t target_id;
int operator()(const SensorTask& node) const
{
if (target_deadline != node.deadline)
{
return (target_deadline < node.deadline) ? -1 : 1;
}
if (target_id != node.id)
{
return (target_id < node.id) ? -1 : 1;
}
return 0;
}
};
uint32_t id;
uint32_t deadline; // next tick at which this task must be serviced
uint32_t period;
};
typedef etl::intrusive_avl_tree<SensorTask> Schedule;
// The only place a task gets linked - called whenever a module attaches,
// or whenever a task is rescheduled after being serviced.
void schedule_task(Schedule& schedule, SensorTask& task)
{
schedule.find_or_insert(SensorTask::Locate{ task.deadline, task.id },
[&task] { return &task; });
}
// Visits every currently-due task in ascending deadline order, services it,
// then reschedules it at its next deadline.
void service_due_tasks(Schedule& schedule, uint32_t now)
{
for (auto it = schedule.begin(); (it != schedule.end()) && (it->deadline <= now); )
{
SensorTask& task = *it;
it = schedule.erase(it); // O(log N), rebalances
std::cout << "servicing task " << task.id << " (was due at " << task.deadline << ")\n";
task.deadline = now + task.period;
schedule_task(schedule, task); // reschedule at the new deadline
}
}
void print_schedule(const Schedule& schedule)
{
for (auto it = schedule.begin(); it != schedule.end(); ++it)
{
std::cout << " task " << it->id << " due at " << it->deadline << "\n";
}
}
int main()
{
Schedule schedule;
// Driver registry: a plain std::vector that a subsystem uses to own its
// tasks for whatever reason it already needs to - the tree does not
// require this vector to exist, it's just a stand-in for "wherever the
// real owner happens to keep the object".
std::vector<SensorTask> driver_registry;
driver_registry.reserve(2);
driver_registry.emplace_back(1, 100, 1000);
schedule_task(schedule, driver_registry.back());
driver_registry.emplace_back(2, 300, 1000);
schedule_task(schedule, driver_registry.back());
std::cout << "Initial schedule (ascending deadline):\n";
print_schedule(schedule);
std::cout << "\nschedule.size() before growth: " << schedule.size() << "\n";
// driver_registry.reserve()'d capacity was 2; this emplace_back forces a
// reallocation. std::vector move-constructs every existing SensorTask
// into the new storage. Each move steals its task's exact tree position
// (O(1), no rebalancing) - the tree needs no code of its own to survive
// the underlying storage moving.
driver_registry.emplace_back(3, 200, 1100);
schedule_task(schedule, driver_registry.back());
std::cout << "schedule.size() after growth: " << schedule.size() << "\n\n";
std::cout << "Schedule after growth (still correctly ordered):\n";
print_schedule(schedule);
std::cout << "\nServicing due tasks at tick 250:\n";
service_due_tasks(schedule, 250);
std::cout << "\nSchedule after servicing:\n";
print_schedule(schedule);
// Unplugging a module removes its SensorTask from driver_registry. Task 2
// isn't the last element, so std::vector::erase() doesn't just destroy it
// in place: it move-assigns the following element (task 3) into task 2's
// slot, then destroys the now-duplicate last slot. That move-assignment
// is SensorTask's (implicitly-generated) one, which calls
// link_type::operator=(link_type&&): it first unlinks *this* - task 2's
// slot, still linked at deadline 300 - from the tree, then takes over
// task 3's tree position. The destructor that runs afterwards, on the
// vacated last slot, fires on an already-unlinked, moved-from object, so
// it has nothing left to do - no explicit schedule.erase() call is made
// here, and no dangling entry is left behind.
std::cout << "\nschedule.size() before unplugging task 2: " << schedule.size() << "\n";
auto pos = std::find_if(driver_registry.begin(), driver_registry.end(),
[](const SensorTask& t) { return t.id == 2; });
driver_registry.erase(pos); // move-assigns task 3 over task 2, unlinking task 2
std::cout << "schedule.size() after unplugging task 2: " << schedule.size() << "\n";
return 0;
}
```
## Program output
```
Initial schedule (ascending deadline):
task 1 due at 100
task 2 due at 300
schedule.size() before growth: 2
schedule.size() after growth: 3
Schedule after growth (still correctly ordered):
task 1 due at 100
task 3 due at 200
task 2 due at 300
Servicing due tasks at tick 250:
servicing task 1 (was due at 100)
servicing task 3 (was due at 200)
Schedule after servicing:
task 2 due at 300
task 1 due at 1250
task 3 due at 1350
schedule.size() before unplugging task 2: 3
schedule.size() after unplugging task 2: 2
```

2
docs/utilities/intrusive-links.md
View File

@ -297,6 +297,8 @@ template <const size_t ID_>
struct tree_link
```
`etl::intrusive_avl_tree` is built on top of `tree_link` - see [intrusive_avl_tree]({{% relref "../containers/trees/intrusive-avl-tree.md" %}}).
### Template parameters
```cpp
const size_t ID_

76
hugo/assets/css/custom.css
View File

@ -420,4 +420,80 @@ tr td code{
font-size: 0.8rem !important;
}
.complexity {
padding: 3px 6px;
border-radius: 999px;
font-weight: 600;
letter-spacing: 0.04em;
border: 1px solid transparent;
}
html.dark .complexity {
&.constant {
color: #7CFFB2;
background: rgba(0,255,120,0.08);
border-color: rgba(0,255,120,0.25);
}
&.logarithmic {
color: #7CE7FF;
background: rgba(0,200,255,0.08);
border-color: rgba(0,200,255,0.18);
}
&.linear {
color: #6AB7FF;
background: rgba(80,140,255,0.08);
border-color: rgba(80,140,255,0.18);
}
&.linearithmic {
color: #B78CFF;
background: rgba(140,80,255,0.08);
border-color: rgba(140,80,255,0.18);
}
&.polynomial {
color: #FFD36A;
background: rgba(255,190,0,0.08);
border-color: rgba(255,190,0,0.18);
}
&.exponential {
color: #ff7d9c;
background: rgba(255,0,80,0.08);
border-color: rgba(255,0,80,0.2);
}
}
html:not(.dark) .complexity {
&.constant {
color: #15803D;
background: rgba(34,197,94,0.10);
border-color: rgba(34,197,94,0.22);
}
&.logarithmic {
color: #0E7490;
background: rgba(6,182,212,0.10);
border-color: rgba(6,182,212,0.22);
}
&.linear {
color: #2563EB;
background: rgba(59,130,246,0.10);
border-color: rgba(59,130,246,0.22);
}
&.linearithmic {
color: #7C3AED;
background: rgba(139,92,246,0.10);
border-color: rgba(139,92,246,0.22);
}
&.polynomial {
color: #CA8A04;
background: rgba(234,179,8,0.12);
border-color: rgba(234,179,8,0.24);
}
&.exponential {
color: #DC2626;
background: rgba(239,68,68,0.10);
border-color: rgba(239,68,68,0.22);
}
}

3
hugo/layouts/shortcodes/complexity.html Normal file
View File

@ -0,0 +1,3 @@
<span class="complexity {{ .Get 0 }}">
𝐎({{ .Get 1 }})
</span>