PR for Modernizing for C++11: replacing instances of SmallVector #6817

M3Henry · Jun 6, 2018

After discussions on IRC about modernising the code to make it easier to work on, replacing non time-critical instances of SmallVector<T,S> with std::vector was deemed more practical than making SmallVector STL-like.

This allows use of range-based for loops and STL algorithms to improve expressiveness in the code base.

No changes to functionality have been made.

#6649

LordAro · Jun 6, 2018

Some build system changes will be needed to enforce c++11/14

nielsmh · Jun 23, 2018

LinkGraphJob::NodeAnnotationVector and GUIBridgeList (in bridge_gui.cpp) are both vectors of non-POD data types so those two are the real important ones to fix.

The former does get fixed by these changes, but GUIBridgeList does not, and from my cursory check it might be more complex to fix that one. The reason the contained data is non-POD is because of the inclusion of Money fields, that type has custom constructors which makes it non-POD.

TrueBrain · Jul 22, 2018

TrueBrain reviewed Jul 22, 2018

View changes

Sorry for the slow responses; we have not forgotten about this PR. I did a fast pass on the code, and "it looks fine". The main issue with patches like this is of course that reviewing takes a bit of time ;)

As mentioned earlier, it would also need some changes to the configure system, as it should detect if the used compiler can handle C++11, and warn if not. Otherwise we will have a bug-tracker full of the same error in no time :)

So please bare with us while we prepare for modern standards ;)

TrueBrain · Jul 22, 2018

src/network/core/host.cpp

@@ -102,7 +102,18 @@ static void NetworkFindBroadcastIPsInternal(NetworkAddressList *broadcast) // GE
 		if (ifa->ifa_broadaddr->sa_family != AF_INET) continue;

 		NetworkAddress addr(ifa->ifa_broadaddr, sizeof(sockaddr));
-		if (!broadcast->Contains(addr)) *broadcast->Append() = addr;
+		//if (!broadcast->Contains(addr)) *broadcast->Append() = addr;


I feel that this was by accident?

The removal was intentional, lines 108-116 replace it, but I shouldn't have left it commented out.
TBH the following would be much better than a hand rolled loop.

if (broadcast->end() == std::find(broadcast->begin(), broadcast->end(), addr) broadcast->push_back(addr);

M3Henry · Jul 27, 2018

@TrueBrain If the intention is to keep modernizing the code, then I can go over this patch series and do it properly. Looking back on this, it doesn't seem I put forward my best.

LordAro · Sep 19, 2018

LordAro requested changes Sep 19, 2018

View changes

First pass at an actual review, mostly codestyle related.

Firstly, great job on actually getting this done and working - it's a huge amount of effort (I tried SmallMap once, I gave up pretty quickly)

However, if I'm honest, I'm not sure this will ever get merged in its current form. There are just too many separate changes all bundled together. It would be better if it were a separate commit per change, rather than per file/type. Something like a commit to add new adapter functions, then one that actually converts types & functions (huge commit, but fundamentally quite simple to read through), then followed by a few commits that "neaten" usage (for loops, auto, etc)

Couple general comments:

Usage of and, or & not instead of &&, || & !
Lots of uses of auto here. Its desired usage in OTTD is yet to be formalised, but IMO it should only be used when the type of the variable doesn't matter, or if the typename has already been stated somewhere nearby. There's a few places in which that's not happened

LordAro · Sep 19, 2018

src/core/smallvec_type.hpp

@@ -7,14 +7,131 @@
 * See the GNU General Public License for more details. You should have received a copy of the GNU General Public License along with OpenTTD. If not, see <http://www.gnu.org/licenses/>.
 */

-/** @file smallvec_type.hpp Simple vector class that allows allocating an item without the need to copy this->data needlessly. */
+/** @file smallvec_type.hpp Simple vector class that allows allocating an item without the need to copy this->data needlessly.


text should be on a separate line, to preserve alignment

(applies to lots of functions)

LordAro · Sep 19, 2018

src/core/smallvec_type.hpp

+ * @tparam T the contained type of the vector
+ */
+template <typename T>
+typename std::vector<T>::const_iterator Find(const std::vector<T>& vec, const T& elem) {


opening brace of a function should be on a newline

(applies to various functions)

LordAro · Sep 19, 2018

src/core/smallvec_type.hpp

+ * @param vec the vector to be searched
+ * @param elem the element to search for
+ * @tparam T the contained type of the vector
+ */


missing @return doxygen command

(applies to various functions)

LordAro · Sep 19, 2018

src/economy.cpp

@@ -77,7 +77,7 @@ static inline int32 BigMulS(const int32 a, const int32 b, const uint8 shift)
 	return (int32)((int64)a * (int64)b >> shift);
 }

-typedef SmallVector<Industry *, 16> SmallIndustryList;
+using SmallIndustryList = std::vector<Industry *> ;


extra space before ;

LordAro · Sep 19, 2018

src/fios.cpp

@@ -397,7 +397,7 @@ static void FiosGetFileList(SaveLoadOperation fop, fios_getlist_callback_proc *c
 	{
 		SortingBits order = _savegame_sort_order;
 		_savegame_sort_order = SORT_BY_NAME | SORT_ASCENDING;
-		QSortT(file_list.files.Begin(), file_list.files.Length(), CompareFiosItems);
+		QSortT(file_list.files.data(), file_list.Length(), CompareFiosItems);


It's a bit more work involved, but since you're converting some stuff in this PR other than just converting SmallVector, perhaps converting to use std::sort would be a good idea as well

Alternatively, it should be a separate PR (along with the various other changes in this PR) :)

I agree that the standard algorithms would be ideal where applicable, a seperate PR would make most sense.

LordAro · Sep 19, 2018

src/network/network_content.h

 	/** Get the end of the content inf iterator. */
-	ConstContentIterator End() const { return this->infos.End(); }
+	ConstContentIterator End() const { return this->infos.end(); }


This looks like another class that could be refactored to be "better", instead of reimplementing a load of std::vector functions

LordAro · Sep 19, 2018

src/station_gui.cpp

 		}
+		else ++ts;


braces required here

LordAro · Sep 19, 2018

src/texteff.cpp

-	for (i = 0; i < _text_effects.Length(); i++) {
-		if (_text_effects[i].string_id == INVALID_STRING_ID) break;
+	auto it = _text_effects.begin();
+	for (; it < _text_effects.end(); ++it) {


could probably be merged onto a single line

it needed to persist beyond the scope of the for loop. However, this is basically std::find_if, so I would re-write it with that instead.

LordAro · Sep 19, 2018

src/timetable_cmd.cpp

 			w->lateness_counter = 0;
 			ClrBit(w->vehicle_flags, VF_TIMETABLE_STARTED);
 			/* Do multiplication, then division to reduce rounding errors. */
-			w->timetable_start = start_date + idx * total_duration / num_vehs / DAY_TICKS;
+			w->timetable_start = start_date + idx++ * total_duration / num_vehs / DAY_TICKS;


idx++ in the middle is a bit scary, should probably be in a separate statement

LordAro · Sep 19, 2018

src/core/smallvec_type.hpp

+ * @tparam T the contained type of the vector
+ */
+template <typename T>
+void Reset(std::vector<T>& vec) {


pretty sure this function isn't necessary - can be replaced with a clear() - the capacity reduction was just due to how the allocations worked

It may be a good idea to keep the capacity reduction where it is used in case that behavior is desired.
According to cppreference: "Leaves the capacity() of the vector unchanged"

M3Henry · Sep 19, 2018

@LordAro
Rather than doing once large PR for a bunch of types at once, would it make more sense to submit one PR per type and do a proper conversion using standard algorithms with each one? (To avoid a PR flood I'd submit the next PR once the previous one is accepted)

TrueBrain · Jan 5, 2019

We recently switched from Jenkins as CI to Azure Pipelines as CI. This means you need to rebase before this Pull Request will pass its checks. Sorry for the troubles! (might not be applicable to this PR)

M3Henry force-pushed the M3Henry:c++11 branch from edde5ab to 072d1b9 Jun 6, 2018

TrueBrain added the needs-review label Jul 19, 2018

M3Henry closed this Jul 27, 2018

M3Henry reopened this Jul 27, 2018

LordAro added waiting-on-author and removed needs-review labels Sep 19, 2018

LordAro added the wip label Oct 27, 2018

LordAro referenced this pull request Jan 5, 2019
Closed
Modernizing for C++11: replacing instances of SmallVector #6649

andythenorth added question stale labels Jan 6, 2019

OpenTTD/OpenTTD

Join GitHub today

PR for Modernizing for C++11: replacing instances of SmallVector #6817

Conversation

M3Henry commented Jun 6, 2018 • edited

M3Henry added some commits Dec 30, 2017

M3Henry force-pushed the M3Henry:c++11 branch from edde5ab to 072d1b9 Jun 6, 2018

This comment has been minimized.

LordAro commented Jun 6, 2018

This comment has been minimized.

nielsmh commented Jun 23, 2018 • edited

TrueBrain added the needs-review label Jul 19, 2018

TrueBrain reviewed Jul 22, 2018 View changes

TrueBrain left a comment

This comment has been minimized.

This comment has been minimized.

This comment has been minimized.

M3Henry commented Jul 27, 2018

M3Henry closed this Jul 27, 2018

M3Henry reopened this Jul 27, 2018

LordAro requested changes Sep 19, 2018 View changes

LordAro left a comment

This comment has been minimized.

This comment has been minimized.

This comment has been minimized.

This comment has been minimized.

This comment has been minimized.

This comment has been minimized.

M3Henry Sep 19, 2018 • edited

This comment has been minimized.

This comment has been minimized.

This comment has been minimized.

This comment has been minimized.

This comment has been minimized.

This comment has been minimized.

This comment has been minimized.

LordAro added waiting-on-author and removed needs-review labels Sep 19, 2018

This comment has been minimized.

M3Henry commented Sep 19, 2018

LordAro added the wip label Oct 27, 2018

LordAro referenced this pull request Jan 5, 2019

This comment has been minimized.

TrueBrain commented Jan 5, 2019

andythenorth added question stale labels Jan 6, 2019

M3Henry commented Jun 6, 2018 •

edited

M3Henry force-pushed the M3Henry:c++11 branch from `edde5ab` to `072d1b9` Jun 6, 2018

nielsmh commented Jun 23, 2018 •

edited

TrueBrain reviewed Jul 22, 2018

View changes

LordAro requested changes Sep 19, 2018

View changes

M3Henry Sep 19, 2018 •

edited