digitalmars.D.learn - use of struct vs class
- mark (31/32) Mar 07 2020 I have this struct (with details omitted ... for brevity):
- drug (5/8) Mar 07 2020 Do you consider using pointers in AA:
- mark (53/61) Mar 07 2020 I've done some changes including using Deb* as you suggested:
- mark (1/1) Mar 07 2020 Instead of deb.clear I'm now doing deb = null;
- mark (31/31) Mar 07 2020 change #1:
- drug (5/43) Mar 07 2020 try
- Steven Schveighoffer (14/52) Mar 07 2020 Hm... I'd say:
- mark (114/114) Mar 07 2020 I've now gone back to using structs direct without pointers but
- Steven Schveighoffer (6/17) Mar 07 2020 RedBlackTree is a class. Make sure you allocate it. The above shows
- mark (24/24) Mar 07 2020 Steve, thank you once again. Now it compiles & runs!
- drug (3/21) Mar 07 2020 You can use predicate to solve the issue like you suggest below. But I'm...
I have this struct (with details omitted ... for brevity):
struct Deb {
string name;
...
RedBlackTree!string tags;
void clear() { name = ""; ...; tags.clear; }
bool valid() { return !(name.empty || description.empty); }
}
I plan to store >65K of these (with potential for growth to
250K) in an AA:
Deb[string] debForName;
I plan to populate debForName by reading data files (actually
Debian Packages files) like this:
Deb deb;
auto file = File(filename):
foreach(line; file.byLine) {
if (line.empty) {
if (deb.valid) // end of package
debForName[deb.name] = deb; // XXX
// else report incomplete package
deb.clear;
continue;
}
... // populate the deb
}
if (deb.valid)
debForName[deb.name] = deb;
I'm assuming that line XXX will copy the Deb including the tree
(which as usual I'm using as a set -- I really miss a set class
in D!). Will this work (I'll find out myself next week when I get
further, but D experts can likely tell from the above).
Should Deb be a class rather than a struct?
Mar 07 2020
07.03.2020 13:20, mark пишет:I have this struct (with details omitted [ snip ] Should Deb be a class rather than a struct?Do you consider using pointers in AA: ``` Deb*[string] debForName; ```
Mar 07 2020
On Saturday, 7 March 2020 at 10:30:06 UTC, drug wrote:07.03.2020 13:20, mark пишет:I've done some changes including using Deb* as you suggested: struct Deb { string name; ... RedBlackTree!string tags; bool valid() { return !(name.empty || description.empty); } size_t toHash() const safe nothrow { return typeid(name).getHash(&name); // names are unique } bool opEquals(const Deb other) const safe pure nothrow { return name == other.name; // names are unique } int opCmp(ref const Deb other) const { return cmp(name, other.name); // names are unique } } Which I now want to store in: RedBlackTree!Deb* debs; // name-ordered list of deb packages And now I'm populating like this: Deb* deb; auto file = File(filename); foreach(line; file.byLine) { line = strip(line); if (line.empty) { if (deb != null && deb.valid) { debs.insert(deb); deb.clear; } // else report incomplete package continue; } if (deb == null) deb = new Deb; ... } if (deb != null && deb.valid) debs.insert(deb); But it crashes with: Performing "debug" build using /home/mark/opt/ldc2-1.20.0-linux-x86_64/bin/ldc2 for x86_64. gtk-d:gtkd 3.9.0: target for configuration "library" is up to date. debfind ~master: building configuration "application"... src/model.d(96,36): Error: template std.container.rbtree.RedBlackTree!(Deb, "a < b", false).RedBlackTree.stableInsert cannot deduce function from argument types !()(Deb*), candidates are: /home/mark/opt/ldc2-1.20.0-linux-x86_64/bin/../import/std/containe /rbtree.d(1256,12): stableInsert(Stuff)(Stuff stuff) with Stuff = Deb* must satisfy the following constraint: isImplicitlyConvertible!(Stuff, Elem) ...I have this struct (with details omitted [ snip ] Should Deb be a class rather than a struct?Do you consider using pointers in AA: ``` Deb*[string] debForName; ```
Mar 07 2020
if (line.empty) {
if (deb != null && deb.valid)
debs.insert(deb);
else // report incomplete
deb = null;
continue;
}
if (deb == null)
deb = new Deb;
bool opEquals(const Deb* other) const safe pure nothrow {
int opCmp(ref const Deb* other) const {
Just changed to pointers & moved to a separate file.
So now I just have one error in line 12 of model.d:
RedBlackTree!Deb* debs; // name-ordered list of deb packages LINE
12
Performing "debug" build using
/home/mark/opt/ldc2-1.20.0-linux-x86_64/bin/ldc2 for x86_64.
gtk-d:gtkd 3.9.0: target for configuration "library" is up to
date.
debfind ~master: building configuration "application"...
src/model.d(12,9): Error: template instance
std.container.rbtree.RedBlackTree!(Deb) does not match template
declaration RedBlackTree(T, alias less = "a < b", bool
allowDuplicates = false)
with T = Deb
must satisfy the following constraint:
is(typeof(binaryFun!less(T.init, T.init)))
/home/mark/opt/ldc2-1.20.0-linux-x86_64/bin/ldc2 failed with exit
code 1.
Mar 07 2020
drug <drug2004 bk.ru>
07.03.2020 13:58, mark пишет:
if (line.empty) {
if (deb != null && deb.valid)
debs.insert(deb);
else // report incomplete
deb = null;
continue;
}
if (deb == null)
deb = new Deb;
bool opEquals(const Deb* other) const safe pure nothrow {
int opCmp(ref const Deb* other) const {
Just changed to pointers & moved to a separate file.
So now I just have one error in line 12 of model.d:
RedBlackTree!Deb* debs; // name-ordered list of deb packages LINE 12
Performing "debug" build using
/home/mark/opt/ldc2-1.20.0-linux-x86_64/bin/ldc2 for x86_64.
gtk-d:gtkd 3.9.0: target for configuration "library" is up to date.
debfind ~master: building configuration "application"...
src/model.d(12,9): Error: template instance
std.container.rbtree.RedBlackTree!(Deb) does not match template
declaration RedBlackTree(T, alias less = "a < b", bool allowDuplicates =
false)
with T = Deb
must satisfy the following constraint:
is(typeof(binaryFun!less(T.init, T.init)))
/home/mark/opt/ldc2-1.20.0-linux-x86_64/bin/ldc2 failed with exit code 1.
try
```
RedBlackTree!(Deb*) debs;
```
Mar 07 2020
On 3/7/20 5:58 AM, mark wrote:
if (line.empty) {
if (deb != null && deb.valid)
debs.insert(deb);
else // report incomplete
deb = null;
continue;
}
if (deb == null)
deb = new Deb;
bool opEquals(const Deb* other) const safe pure nothrow {
int opCmp(ref const Deb* other) const {
Just changed to pointers & moved to a separate file.
So now I just have one error in line 12 of model.d:
RedBlackTree!Deb* debs; // name-ordered list of deb packages LINE 12
Performing "debug" build using
/home/mark/opt/ldc2-1.20.0-linux-x86_64/bin/ldc2 for x86_64.
gtk-d:gtkd 3.9.0: target for configuration "library" is up to date.
debfind ~master: building configuration "application"...
src/model.d(12,9): Error: template instance
std.container.rbtree.RedBlackTree!(Deb) does not match template
declaration RedBlackTree(T, alias less = "a < b", bool allowDuplicates =
false)
with T = Deb
must satisfy the following constraint:
is(typeof(binaryFun!less(T.init, T.init)))
/home/mark/opt/ldc2-1.20.0-linux-x86_64/bin/ldc2 failed with exit code 1.
Hm... I'd say:
1. Don't use a pointer for the element. Just use the struct directly.
Using a pointer is bad because it's now going to compare pointers, and
not the element data. Not only that, but RBNodes are stored as
heap-allocated structs, so you are wasting a lot of memory by allocating
another heap allocated thing to get stored inside there.
2. RedBlackTree allows you to identify the relationship that you
consider unique by providing a "less" function. Instead of instrumenting
your Deb type, which might affect other usages, just do:
RedBlackTree!(Deb, (a, b) => a.name < b.name)
No need to add opCmp and opEquals (if that doesn't make sense in other
contexts).
-Steve
Mar 07 2020
I've now gone back to using structs direct without pointers but
I'm still doing something wrong.
struct Deb {
string name;
...
RedBlackTree!string tags;
bool valid() { return !(name.empty || description.empty); }
void clear() { name = ""; ...; tags.clear; }
}
RedBlackTree!(Deb, (a, b) => a.name < b.name) debs;
Deb deb;
auto file = File(filename);
foreach(line; file.byLine) {
line = strip(line);
if (line.empty) {
if (deb.valid)
debs.insert(deb); // XXX
else // report incomplete
deb.clear;
continue;
}
...
}
if (deb.valid)
debs.insert(deb);
This compiles & crashes with "Program exited with code -11".
What I'm trying to do but clearly don't understand is this:
1. create a struct
repeat:
2.1. populate the struct
2.2. copy the struct into an rbtree // XXX
2.3. clear the original struct
2.4. loop
At XXX does a copy take place? I thought it did with structs?
Anyway, here's the backtrace:
$ gdb DebFind
GNU gdb (Ubuntu 8.1-0ubuntu3.2) 8.1.0.20180409-git
...
Reading symbols from DebFind...done.
(gdb) run
Starting program: /home/mark/app/d/debfind/DebFind
[Thread debugging using libthread_db enabled]
Using host libthread_db library
"/lib/x86_64-linux-gnu/libthread_db.so.1".
[New Thread 0x7ffff6e02700 (LWP 2366)]
...
Thread 1 "DebFind" received signal SIGSEGV, Segmentation fault.
0x0000555555701ef0 in
_D3std9container6rbtree__T12RedBlackTreeTAyaVQea5_61203c2062Vbi0ZQBn
emptyMFNaNbNdNiNfZb (this=0x0)
at
/home/mark/opt/ldc2-1.20.0-linux-x86_64/bin/../import/std/container/rbtree.d:967
967 return _end.left is null;
(gdb) bt
_D3std9container6rbtree__T12RedBlackTreeTAyaVQea5_61203c2062Vbi0ZQBn
emptyMFNaNbNdNiNfZb (this=0x0)
at
/home/mark/opt/ldc2-1.20.0-linux-x86_64/bin/../import/std/container/rbtree.d:967
qtrac.debfind.model.Model.readPackageFile(immutable(char)[])
(this=0x0, filename=...) at model.d:66
warning: (Internal error: pc 0x55555591b874 in read in psymtab,
but not in symtab.)
warning: (Internal error: pc 0x55555591b820 in read in psymtab,
but not in symtab.)
warning: (Internal error: pc 0x55555591b874 in read in psymtab,
but not in symtab.)
qtrac.debfind.model.Model.initialize(int) (this=0x0,
maxDebNamesForWord=100) at model.d:31
warning: (Internal error: pc 0x55555591b874 in read in psymtab,
but not in symtab.)
_D5qtrac7debfind9appwindow9AppWindow6__ctorMFC3gtk11ApplicationQnZCQCnQCkQCfQBy
(this=0x7ffff7ece630, application=0x7ffff7ed1360) at appwindow.d:31
_D5qtrac7debfind3app4mainFAAyaZ__T12__dgliteral2TC3gio11Appl
cationQnZQBkMFQBaZv (GioApplication=0x7ffff7ed1360) at app.d:16
_D7gobject8DClosureQj__T17d_closure_marshalTDFC3gio11ApplicationQnZvZQBtUPSQCv1c5types8GClosurePSQDr
wQw6GValuekQrPvQcZv (closure=0x555555c37690, return_value=0x0,
n_param_values=1, param_values=0x7ffff7ef46e0, invocation_hint=0x7fffffffd710
"\b", marshal_data=0x0) at DClosure.d:122
/usr/lib/x86_64-linux-gnu/libgobject-2.0.so.0
/usr/lib/x86_64-linux-gnu/libgobject-2.0.so.0
/usr/lib/x86_64-linux-gnu/libgobject-2.0.so.0
/usr/lib/x86_64-linux-gnu/libgobject-2.0.so.0
warning: (Internal error: pc 0x55555591b874 in read in psymtab,
but not in symtab.)
/usr/lib/x86_64-linux-gnu/libgio-2.0.so.0
warning: (Internal error: pc 0x55555591b874 in read in psymtab,
but not in symtab.)
/usr/lib/x86_64-linux-gnu/libgio-2.0.so.0
warning: (Internal error: pc 0x55555591b874 in read in psymtab,
but not in symtab.)
0x55555591b874 in read in psymtab, but not in symtab.)
_D3gio11ApplicationQn3runMFAAyaZiwarning: (Internal error: pc
0x55555591b874 in read in psymtab, but not in symtab.)
warning: (Internal error: pc 0x55555591b874 in read in psymtab,
but not in symtab.)
(warning: (Internal error: pc 0x55555591b874 in read in psymtab,
but not in symtab.)
this=0x7ffff7ed1360, warning: (Internal error: pc 0x55555591b874
in read in psymtab, but not in symtab.)
argv=...)warning: (Internal error: pc 0x55555591b874 in read in
psymtab, but not in symtab.)
at Application.dwarning: (Internal error: pc 0x55555591b874 in
read in psymtab, but not in symtab.)
warning: (Internal error: pc 0x55555591b874 in read in psymtab,
but not in symtab.)
:931
And thanks for helping!
Mar 07 2020
On 3/7/20 8:22 AM, mark wrote:0x0000555555701ef0 in _D3std9container6rbtree__T12RedBlackTreeTAyaVQea5_61203c2062Vbi0ZQBn emptyMFNaNbNdNiNfZb (this=0x0) at /home/mark/opt/ldc2-1.20.0-linux-x86_64/bin/../import/std/co tainer/rbtree.d:967 967 return _end.left is null; (gdb) bt _D3std9container6rbtree__T12RedBlackTreeTAyaVQea5_61203c2062Vbi0ZQBn emptyMFNaNbNdNiNfZb (this=0x0)RedBlackTree is a class. Make sure you allocate it. The above shows "this = 0x0". I saw a few other "this = 0x0" items, but perhaps that's OK for your code, I don't know what those types are. -Steve
Mar 07 2020
Steve, thank you once again. Now it compiles & runs! I now create my tree like this: auto debs = new RedBlackTree!(Deb, (a, b) => a.name < b.name); (I feel that the rbtree docs are inadequate regarding creating new empty trees, so have submitted a bug report: https://issues.dlang.org/show_bug.cgi?id=20646 ) The other problem I had was that the Deb I created on the stack had its own internal rbtree that was always null. To address this I changed tags in the Deb struct to this: auto tags = new RedBlackTree!string; so it always creates a new empty tree. And I renamed clear to reset and now do this: void reset() { name = ""; ... tags = new RedBlackTree!string; ... } I assume that I can safely rely on the GC to clean up for me. So now I think I can safely use the pattern (1., repeat 2.1, ... ) described earlier: certainly it builds and runs. Next I'll have to try really populating the Debs including their tag trees and populating the debs tree of Debs structs. Thank you.
Mar 07 2020
07.03.2020 15:58, Steven Schveighoffer пишет:Hm... I'd say: 1. Don't use a pointer for the element. Just use the struct directly. Using a pointer is bad because it's now going to compare pointers, and not the element data. Not only that, but RBNodes are stored as heap-allocated structs, so you are wasting a lot of memory by allocating another heap allocated thing to get stored inside there.You can use predicate to solve the issue like you suggest below. But I'm totally missed that RBNodes are heap-allocated, thanks!2. RedBlackTree allows you to identify the relationship that you consider unique by providing a "less" function. Instead of instrumenting your Deb type, which might affect other usages, just do: RedBlackTree!(Deb, (a, b) => a.name < b.name) > No need to add opCmp and opEquals (if that doesn't make sense in other contexts). -Steve
Mar 07 2020