ltldo: add portfolio options
Fixes #206. * bin/ltldo.cc: Implement --smallest and --greatest. * tests/core/ltldo2.test: Test them. * NEWS, doc/org/ltldo.org: Document them.
This commit is contained in:
Alexandre Duret-Lutz
parent
267f819a9d
commit
a4b575db1c
4 changed files with 237 additions and 18 deletions
|
|
@ -182,12 +182,13 @@ ltldo --help | sed -n '/character sequences:/,/^$/p' | sed '1d;$d'
|
|||
#+END_SRC
|
||||
|
||||
#+RESULTS:
|
||||
: %% a single %
|
||||
: %f,%s,%l,%w the formula as a (quoted) string in Spot, Spin,
|
||||
: LBT, or Wring's syntax
|
||||
: %F,%S,%L,%W the formula as a file in Spot, Spin, LBT, or
|
||||
: Wring's syntax
|
||||
: %O,%D the automaton is output as either (%O) HOA/never
|
||||
: claim/LBTT, or (%D) in LTL2DSTAR's format
|
||||
: %O the automaton output in HOA, never claim, LBTT, or
|
||||
: ltl2dstar's format
|
||||
|
||||
Contrarily to =ltlcross=, it this not mandatory to specify an output
|
||||
filename using one of the sequence for that last line. For instance
|
||||
|
|
@ -464,6 +465,150 @@ State: 0 "accept_init" {0}
|
|||
--END--
|
||||
#+END_SRC
|
||||
|
||||
* Acting as a portfolio of translators
|
||||
:PROPERTIES:
|
||||
:CUSTOM_ID: portfolio
|
||||
:END:
|
||||
|
||||
Here is a formula on which different translators produce Büchi automata of
|
||||
different sizes (states and edges):
|
||||
|
||||
#+BEGIN_SRC sh :results verbatim :exports both
|
||||
ltldo ltl2ba ltl3ba 'ltl2tgba -s' -f 'F(a & Xa | FGa)' \
|
||||
--stats='%T: %s st. (%n non-det.), %e ed.'
|
||||
#+END_SRC
|
||||
|
||||
#+RESULTS:
|
||||
: ltl2ba: 5 st. (2 non-det.), 9 ed.
|
||||
: ltl3ba: 3 st. (1 non-det.), 4 ed.
|
||||
: ltl2tgba -s: 3 st. (0 non-det.), 5 ed.
|
||||
|
||||
Instead of outputting the result of the translation of each formula by each
|
||||
translator, =ltldo= can also be configured to output the smallest
|
||||
automaton obtained for each formula:
|
||||
|
||||
#+BEGIN_SRC sh :results verbatim :exports both
|
||||
ltldo ltl2ba ltl3ba 'ltl2tgba -s' -f 'F(a & Xa | FGa)' --smallest
|
||||
#+END_SRC
|
||||
|
||||
#+RESULTS:
|
||||
#+begin_example
|
||||
HOA: v1
|
||||
States: 3
|
||||
Start: 0
|
||||
AP: 1 "a"
|
||||
acc-name: Buchi
|
||||
Acceptance: 1 Inf(0)
|
||||
properties: trans-labels explicit-labels state-acc
|
||||
--BODY--
|
||||
State: 0
|
||||
[t] 0
|
||||
[0] 1
|
||||
State: 1
|
||||
[0] 2
|
||||
State: 2 {0}
|
||||
[t] 2
|
||||
--END--
|
||||
#+end_example
|
||||
|
||||
Therefore, in practice, =ltldo --smallest trans1 trans2 trans3...=
|
||||
acts like a portfolio of translators, always returning the smallest
|
||||
produced automaton.
|
||||
|
||||
The sorting criterion can be specified using =--smallest= or
|
||||
=--greatest=, optionally followed by a format string with
|
||||
=%=-sequences. The default criterion is =%s,%e=, so the number of
|
||||
states will be compared first, and in case of equality the number of
|
||||
edges. If we desire the automaton that has the fewest states, and in
|
||||
case of equality the smallest number of non-deterministic states, we
|
||||
can use the following command instead.
|
||||
|
||||
#+BEGIN_SRC sh :results verbatim :exports both
|
||||
ltldo ltl2ba ltl3ba 'ltl2tgba -s' -f 'F(a & Xa | FGa)' --smallest=%s,%n
|
||||
#+END_SRC
|
||||
|
||||
#+RESULTS:
|
||||
#+begin_example
|
||||
HOA: v1
|
||||
name: "F(a & Xa)"
|
||||
States: 3
|
||||
Start: 0
|
||||
AP: 1 "a"
|
||||
acc-name: Buchi
|
||||
Acceptance: 1 Inf(0)
|
||||
properties: trans-labels explicit-labels state-acc complete
|
||||
properties: deterministic terminal
|
||||
--BODY--
|
||||
State: 0
|
||||
[!0] 0
|
||||
[0] 1
|
||||
State: 1
|
||||
[!0] 0
|
||||
[0] 2
|
||||
State: 2 {0}
|
||||
[t] 2
|
||||
--END--
|
||||
#+end_example
|
||||
|
||||
We can of course apply this on a large number of formulas. For
|
||||
instance here is a more complex pipeline, where we take 11 patterns
|
||||
from Dwyer et al. (FMSP'98), and print which translator among
|
||||
=ltl2ba=, =ltl3ba=, and =ltl2tgba -s= would produce the smallest
|
||||
automaton.
|
||||
|
||||
#+BEGIN_SRC sh :results verbatim :exports both
|
||||
genltl --dac=10..20 --format=%F:%L,%f |
|
||||
ltldo -F-/2 ltl2ba ltl3ba 'ltl2tgba -s' --smallest --stats='%<,%T'
|
||||
#+END_SRC
|
||||
#+RESULTS:
|
||||
#+begin_example
|
||||
dac-patterns:10,ltl2ba
|
||||
dac-patterns:11,ltl3ba
|
||||
dac-patterns:12,ltl2tgba -s
|
||||
dac-patterns:13,ltl2tgba -s
|
||||
dac-patterns:14,ltl2tgba -s
|
||||
dac-patterns:15,ltl2tgba -s
|
||||
dac-patterns:16,ltl2ba
|
||||
dac-patterns:17,ltl2tgba -s
|
||||
dac-patterns:18,ltl2ba
|
||||
dac-patterns:19,ltl3ba
|
||||
dac-patterns:20,ltl2ba
|
||||
#+end_example
|
||||
|
||||
Note that in case of equality, only the first translator is returned.
|
||||
So when =ltl2ba= is output above, it could be the case that =ltl3ba=
|
||||
or =ltl2tgba -s= are also producing automata of equal size.
|
||||
|
||||
To understand the above pipeline, remove the =ltldo= invocation. The
|
||||
[[file:genltl.org][=genltl=]] command outputs this:
|
||||
|
||||
#+BEGIN_SRC sh :results verbatim :exports both
|
||||
genltl --dac=10..20 --format=%F:%L,%f
|
||||
#+END_SRC
|
||||
#+RESULTS:
|
||||
#+begin_example
|
||||
dac-patterns:10,G((p0 & !p1) -> (!p1 U (!p1 & p2)))
|
||||
dac-patterns:11,!p0 W (p0 W (!p0 W (p0 W G!p0)))
|
||||
dac-patterns:12,Fp0 -> ((!p0 & !p1) U (p0 | ((!p0 & p1) U (p0 | ((!p0 & !p1) U (p0 | ((!p0 & p1) U (p0 | (!p1 U p0)))))))))
|
||||
dac-patterns:13,Fp0 -> (!p0 U (p0 & (!p1 W (p1 W (!p1 W (p1 W G!p1))))))
|
||||
dac-patterns:14,G((p0 & Fp1) -> ((!p1 & !p2) U (p1 | ((!p1 & p2) U (p1 | ((!p1 & !p2) U (p1 | ((!p1 & p2) U (p1 | (!p2 U p1))))))))))
|
||||
dac-patterns:15,G(p0 -> ((!p1 & !p2) U (p2 | ((p1 & !p2) U (p2 | ((!p1 & !p2) U (p2 | ((p1 & !p2) U (p2 | (!p1 W p2) | Gp1)))))))))
|
||||
dac-patterns:16,Gp0
|
||||
dac-patterns:17,Fp0 -> (p1 U p0)
|
||||
dac-patterns:18,G(p0 -> Gp1)
|
||||
dac-patterns:19,G((p0 & !p1 & Fp1) -> (p2 U p1))
|
||||
dac-patterns:20,G((p0 & !p1) -> (p2 W p1))
|
||||
#+end_example
|
||||
|
||||
This is a two-column CSV file where each line is a description of the
|
||||
origin of the formula (=%F:%L=), followed by the formula itself
|
||||
(=%f=). The =ltldo= from the previous pipeline simply takes its input
|
||||
from the second column of its standard input (=-F-/2=), run that
|
||||
formula through the three translator, pick the smallest automaton
|
||||
(=--smallest=), and for this automaton, it display the translator that
|
||||
was used (=%T=) along with the portion of the CSV file that was before
|
||||
the input column (=%<=).
|
||||
|
||||
* Controlling and measuring time
|
||||
|
||||
The run time of each command can be restricted with the =-T NUM=
|
||||
|
|
|
|||
Loading…
Add table
Add a link
Reference in a new issue