Refactor types and names #299

Adda0 · 2023-08-15T06:59:12Z

This PR refactors some types, modifies some names, and in general resolves the majority of the issues from #292 and #283.

I suggest reviewing the code commit after commit in order to get an idea of what each change does and what is the impact of that change.

Further changes regarding transitions iterator and alphabet will come in their respective PRs as those are rather large modifications.

codecov · 2023-08-15T07:06:58Z

Codecov Report

Patch coverage: 84.13% and project coverage change: +0.29% 🎉

Comparison is base (0259d79) 82.58% compared to head (a539cc8) 82.88%.
Report is 2 commits behind head on devel.

Additional details and impacted files

@@            Coverage Diff             @@
##            devel     #299      +/-   ##
==========================================
+ Coverage   82.58%   82.88%   +0.29%     
==========================================
  Files          51       51              
  Lines       10599    10526      -73     
  Branches     1063     1034      -29     
==========================================
- Hits         8753     8724      -29     
+ Misses       1443     1414      -29     
+ Partials      403      388      -15

Files Changed	Coverage Δ
include/mata/nfa/nfa.hh	`89.28% <ø> (-0.37%)`	⬇️
include/mata/nfa/plumbing.hh	`75.00% <ø> (ø)`
include/mata/utils/sparse-set.hh	`76.56% <ø> (+0.75%)`	⬆️
include/mata/utils/util.hh	`82.81% <ø> (-1.68%)`	⬇️
src/strings/nfa-segmentation.cc	`93.42% <0.00%> (ø)`
tests/nfa/nfa-profiling.cc	`0.00% <0.00%> (ø)`
src/alphabet.cc	`53.19% <53.12%> (ø)`
src/strings/nfa-noodlification.cc	`71.49% <60.00%> (+0.06%)`	⬆️
src/nfa/nfa.cc	`59.68% <75.00%> (-2.40%)`	⬇️
tests/strings/nfa-string-solving.cc	`90.50% <75.00%> (ø)`
... and 12 more

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

tfiedor

No major issues.

tfiedor · 2023-08-15T20:18:52Z

include/mata/alphabet.hh

@@ -26,6 +26,8 @@
 namespace Mata {

 using Symbol = unsigned;
+using Word = std::vector<Symbol>;


I understand, you probably discussed this with Lukas, but are you sure it is good to introduce new types, after we aggresively removed lots of types? :D

We did discuss it. The plan is that Word and WordName are to be the interface for working with alphabets. And therefore deserve a type of their own. The user will work with word names (a sequence of strings representing names of each symbol in a word), while Mata works internally with words (a sequence of Mata::Symbols). This is the place to discuss this idea with all of us, so feel free to give your opinion.

I thing having this one named is good, it appears on many places later.

We will keep it, at least until we see how well it works with refactored alphabets (coming soon to you Mata clone!).

tfiedor · 2023-08-15T20:24:13Z

include/mata/nfa/nfa.hh

+     * @param[out] state_renaming Mapping of trimmed states to new states.
+     * @return New @c Nfa instace from @c this with trimmed states.
+     */
+    Nfa trimmed(StateRenaming* state_renaming = nullptr) const { return Nfa{ *this }.trim(state_renaming); }


I think this naming is confusing, having trim and trimmed will only confuse users, so I suggest to brainstorm some better name.

The idea is that the operations that work in-place will usually have the name after the operation (usually a verb) and what it does to the instance the operation is called on. The operations returning a new instance will be named after the result, the new instance, which in this case is the trimmed NFA. Again, this is the idea. Let us discuss this here. trim() and trimmed() are the example of how it will look like. If we agree on using this pattern, all other functions will be renamed similarly.

I get the idea, but the naming is probably not ideal. I though of copy_trimmed() or clone_trimmed() to show that a copy of the original automaton is made.

I think, we should more carefully unify naming convetion (mainly) for: 1. member operations (like nfa.trim(), 2. non-member operation (like res = trim(nfa), 3. member operation returning clone/copy res = nfa.trim().

I though of copy_trimmed() or clone_trimmed() to show that a copy of the original automaton is made.

I would prefer something like this, too, but others have to agree as well.

unify naming convetion

That is what I start doing here. This is an example of how it will look with trim(), other function will follow. The naming convention (as of now, without your suggestion) should be:

trim()

trimmed()

trimmed()

However, a lot of the non-member operations should either disappear and be replaced with their member variants: determinize(nfa) to nfa.determinized(), etc. In some cases, especially for binary operations, both variants will exist: intersection(nfa, nfa2) and nfa.intersection_with(nfa2) plus the in-place variant may exist, too: concanation(nfa, nfa2), nfa.concatenate_with(nfa) (in-place) and nfa.concatenated_with(nfa2).

But this approach still has its faults. We should discuss this further, think about possible problems, ... For further discussion, see #172 and #277 (comment) (and in other discussions in this PR).

I would refactor the trims first, we need to measure the speed and keep only the fastest, and then think about names. This verb /adjective naming system is not used consitently anyway, so it looks even weirder here.
Actually I confused it with revert, which I wanted to evaluate.
Here, lets just keep the in place trim and rename it then.

But we can ad it in a separate pull request, lets merge this one quick.

That can be done. I will remove trimmed() for now.

src/nfa/nfa.cc

include/mata/alphabet.hh

kilohsakul

I commented few small things.

include/mata/alphabet.hh

include/mata/nfa/nfa.hh

kilohsakul · 2023-08-18T07:49:02Z

include/mata/nfa/nfa.hh

+     * @param[out] state_renaming Mapping of trimmed states to new states.
+     * @return New @c Nfa instace from @c this with trimmed states.
+     */
+    Nfa trimmed(StateRenaming* state_renaming = nullptr) const { return Nfa{ *this }.trim(state_renaming); }


I would refactor the trims first, we need to measure the speed and keep only the fastest, and then think about names. This verb /adjective naming system is not used consitently anyway, so it looks even weirder here.
Actually I confused it with revert, which I wanted to evaluate.
Here, lets just keep the in place trim and rename it then.

include/mata/nfa/strings.hh

src/nfa/nfa.cc

kilohsakul · 2023-08-18T08:16:11Z

tests/nfa/nfa.cc

    Nfa aut{20};
    FILL_WITH_AUT_A(aut);
    aut.delta.add(0, EPSILON, 3);
    aut.delta.add(3, EPSILON, 3);
    aut.delta.add(3, EPSILON, 4);

-    auto state_eps_trans{ aut.get_epsilon_transitions(0) };
+    auto state_eps_trans{ aut.delta.epsilon_symbol_posts(0) };


_symbol remove?

Actually, this is not meant to be epsilon_symbol and posts, but epsilon and sybol_posts, as in, SymbolPost for epsilon from a given source. Therefore, the name should be kept as it is.

kilohsakul · 2023-08-18T08:17:37Z

include/mata/alphabet.hh

@@ -26,6 +26,8 @@
 namespace Mata {

 using Symbol = unsigned;
+using Word = std::vector<Symbol>;


I thing having this one named is good, it appears on many places later.

kilohsakul · 2023-08-18T08:18:39Z

include/mata/nfa/nfa.hh

+     * @param[out] state_renaming Mapping of trimmed states to new states.
+     * @return New @c Nfa instace from @c this with trimmed states.
+     */
+    Nfa trimmed(StateRenaming* state_renaming = nullptr) const { return Nfa{ *this }.trim(state_renaming); }


But we can ad it in a separate pull request, lets merge this one quick.

…in()'

…ol_states()'

…tates

Adda0 · 2023-08-18T12:43:07Z

The changes from reviews that were not applied immediately will be applied in future PRs, namely in alphabet refactorization.

Adda0 added 4 commits August 11, 2023 12:32

Remove duplicated type declaration

1e58fec

Refactor Noodle types

3a6e6a8

Refactor Word types

4c64919

Remove reduce parameter trim_input

5b39918

Allow chaining of trim()

f3cc568

Adda0 force-pushed the refactor_types_and_names branch from d0590b7 to d0b1e8b Compare August 15, 2023 07:36

Adda0 requested review from kilohsakul, tfiedor and jurajsic August 15, 2023 07:37

Adda0 marked this pull request as ready for review August 15, 2023 07:40

tfiedor approved these changes Aug 15, 2023

View reviewed changes

jurajsic approved these changes Aug 16, 2023

View reviewed changes

kilohsakul reviewed Aug 18, 2023

View reviewed changes

include/mata/alphabet.hh Show resolved Hide resolved

kilohsakul reviewed Aug 18, 2023

View reviewed changes

Adda0 added 12 commits August 18, 2023 10:30

Add method 'OrdVector::contains()'

e4ddd76

Merge both 'compute_reachability()' functions into 'reachable_states_…

e17b882

…in()'

Refactor 'Nfa::Nfa::get_epsilon_transitons()' as 'Delta::epsilon_symb…

3741392

…ol_states()'

Use conventional suffix '_' for all private data members

d4e7557

Use iterator over transitions to simplify code

3c1158c

Rename Delta::transform to Delta::renumber_targets

69e0f3b

Refactor 'Delta::add()' to use 'target' and 'source' for transition s…

1fb1cf8

…tates

Test new operations in Delta

ea07cb3

Remove deprecated trim and useful states functions

fc9eec0

Improve comments for BoolVector

20bebfc

Use 'from' instead of 'for' when talking about shortest words

364625e

Remove redundant suffix '_in'

a539cc8

Adda0 force-pushed the refactor_types_and_names branch from d0b1e8b to a539cc8 Compare August 18, 2023 10:27

Adda0 merged commit 1b7128e into devel Aug 18, 2023

Adda0 deleted the refactor_types_and_names branch August 18, 2023 10:38

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Refactor types and names #299

Refactor types and names #299

Adda0 commented Aug 15, 2023

codecov bot commented Aug 15, 2023 •

edited

Loading

tfiedor left a comment

tfiedor Aug 15, 2023

Adda0 Aug 16, 2023

kilohsakul Aug 18, 2023

Adda0 Aug 18, 2023

tfiedor Aug 15, 2023

Adda0 Aug 16, 2023

tfiedor Aug 16, 2023 •

edited

Loading

Adda0 Aug 16, 2023 •

edited

Loading

kilohsakul Aug 18, 2023

kilohsakul Aug 18, 2023 •

edited

Loading

Adda0 Aug 18, 2023

kilohsakul left a comment

kilohsakul Aug 18, 2023

kilohsakul Aug 18, 2023

Adda0 Aug 18, 2023

Adda0 Aug 18, 2023

kilohsakul Aug 18, 2023

kilohsakul Aug 18, 2023 •

edited

Loading

Adda0 commented Aug 18, 2023

Refactor types and names #299

Refactor types and names #299

Conversation

Adda0 commented Aug 15, 2023

codecov bot commented Aug 15, 2023 • edited Loading

Codecov Report

tfiedor left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

tfiedor Aug 16, 2023 • edited Loading

Choose a reason for hiding this comment

Adda0 Aug 16, 2023 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

kilohsakul Aug 18, 2023 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

kilohsakul left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

kilohsakul Aug 18, 2023 • edited Loading

Choose a reason for hiding this comment

Adda0 commented Aug 18, 2023

codecov bot commented Aug 15, 2023 •

edited

Loading

tfiedor Aug 16, 2023 •

edited

Loading

Adda0 Aug 16, 2023 •

edited

Loading

kilohsakul Aug 18, 2023 •

edited

Loading

kilohsakul Aug 18, 2023 •

edited

Loading