Merge branch 'feature-a/release' into 'dev'

Release 2020.06

See merge request !727

algo/release

-2020.05
+2020.06

doc/.gitlab-ci.yml

@@ -16,6 +16,7 @@ build_doc:
       # branch name should contain doc before the first /
       # eg.: doc/blabla or feature-adoc/blabla
       - /^feature-.*a.*\/release$/
+  allow_failure: true
   tags:
     - cidocker
 

doc/CHANGELOG

 
 This changelog concerns vijil-algo, the algorithmic part (C++) of the Vidjil platform.
 
+2020-06-12  The Vidjil Team
+	* New report of 'start of V' and 'end of J' positions #2138, improving the productivity estimation #2142
+	  This is based on a streamlined alignment for reverse sequences (core/segment.cpp) !706
+	* Faster and improved probability computations (core/proba.cpp) !679
+	  This will bring further speed-ups in a future release
+	* Ctrl-C now gracefully stops execution with a warning W09 and some output #4286
+	* Streamlined handling of warnings #3360, also reported with -c designations #3811
+	* Better and more reproducible speed and memory benchmarks (benchmark-releases.py, #4275) #918, new profiling #918
+	* Adoption of 'detected' term, renamed output files to '.detected.vdj.fa' #3413
+	* Updated documentation
+	* New and updated tests
+	* Better release process through a merge request template !685
+
 2020-05-02  The Vidjil Team
 	* Bug closed (crash in very rare cases #4272)
 

doc/vidjil-algo.md

-# vidjil-algo 2020.05
+# vidjil-algo 2020.06
 **Command-line manual**
 
 *The Vidjil team (Mathieu, Mikaël, Aurélien, Florian, Marc, Tatiana and Rayan)*
@@ -496,11 +496,12 @@ as provided by `make germline`.
 The CDR3/JUNCTION detection won't work with custom non-gapped V/J repertoires.
 
 CDR3 are reported as *productive* when they come from an in-frame recombination
-and when the full sequence does not contain any in-frame stop codons.
+and when the sequence does not contain any in-frame stop codons.
 Note that some other software only consider stop codons in the CDR3,
-and may thus under-estimate non-productivity. Vidjil-algo looks for in-frame stop codons
-on all the available sequence (and may sometimes over-estimate non-productivity when
-the sequence contains intronic regions).
+and may thus under-estimate non-productivity.
+When the sequence is long enough to start before the start of the V gene
+or to end after the end of the J gene, vidjil-algo do not consider these intronic sequences
+in the productivity estimation.
 
 The advanced `--analysis-cost` option sets the parameters used in the comparisons between
 the clone sequence and the V(D)J germline genes. The default values should work.