This talk was presented by Tom Mens and Loup Meurice at the EvoLille'15 European Open Symposium on Empirical Software Engineering (December 2015). We report on the evolution of open source Java projects relying on relational database technologies. We empirically analysed how the use of Java database access technologies evolves over time. We carried out a coarse-grained source-code analysis over several thousands of Java projects on GitHub, and complemented this by a fine-grained longitudinal analysis of the co-evolution between database schema changes and source code changes within three large Java projects (OSCAR, OpenEMR and BroadLeaf). The presented results are a first step towards a recommendation system supporting developers in writing database-centered code. The research was part of an interuniversity research project between University of Mons and University of Namur, financed by the F.R.S.-FNRS, Belgium.

1. Evolu&on of database access technologies
in Java-based so6ware projects
Loup Meurice
Anthony Cleve
Csaba Nagy
University of Namur, Belgium
h?p://informa&que.umons.ac.be/genlog/projects/disse
EOSESE 2015 — Lille, France, 3 December 2015
Tom Mens
Mathieu Goeminne
Alexandre Decan
University of Mons, Belgium

2. EOSESE – Lille, France, 3 December 2015
• Context: Interuniversity FNRS-funded research
project between Université de Mons
and Université de Namur
• Focus: Empirical analysis of co-evolu&on between program
code and database (schema) in data-intensive so6ware
systems
• Goal: Provide be?er support for such co-evolu&on to
so6ware developers and project managers
• This presenta&on: Study usage and evolu&on of rela&onal
database access (RDBA) technologies for open source Java
systems
Data-Intensive So6ware
System Evolu&on

3. Focus
3
• Java RDBA technologies

4. 4
Focus
• Java RDBA technologies

5. EOSESE – Lille, France, 3 December 2015 5
1 < h i b e r n a t e mapping>
2 < c l a s s name=" Customer " t a b l e =" AppCustomers ">
3 < id name=" id " type =" i n t " column=" id " / >
4 < p r o p e r t y name="name" column="name" type =" S t r i n g " / >
5 < / c l a s s >
6 < / h i b e r n a t e mapping>
Example of using JPA annotations:
1 @Entity
2 @Table ( name=" AppCustomers " )
3 public c l a s s Customer {
4 @Id
5 private i n t id ;
6 S t r i n g name ;
7 [ . . . ]
8 }
Example of using Hibernate Query Language (HQL):
1 public List <Customer > findAllCustomers ( ) {
2 S t r i n g hql = " s e l e c t c from Customer c " ;
3 return executeQuery ( hql ) ;
4 }
5 public L i s t executeQuery ( S t r i n g hql ) {
6 return s e s s i o n . createQuery ( hql ) . l i s t ( ) ;
7 }
8 public Customer findCustomer ( I n t e g e r id ) {
9 return ( Customer ) s e s s i o n . get ( Customer . class , id ) ;
10 }
Example of direct object manipulation using ActiveJDBC:
1 List <Customer > customers = Customer . where ( "name = ’ John Doe ’ " ) ;
2 Customer john = customers . get ( 0 ) ;
3 john . setName ( " John Smith " ) ;
4 john . s a v e I t ( ) ;
Figure 2: Four examples of Java ORM usage.
6
The mapping can take many different forms, as illustrated in Figure 2. The first example shows
the use of Hibernate configuration files (i.e., .hbm.xml). The second example illustrates the
use of Java annotations based on JPA, the standard API in Java for ORM and data persistence
management. Such a mapping may allow direct operations on objects, attributes and relationships
instead of tables and columns. This is commonly referred as the active record pattern.
Some ORM libraries also provide SQL-inspired languages that allow to write SQL-like queries
using the mappings defined before. The third example in Figure 2 illustrates the Hibernate Query
Language (HQL), and the fourth example uses the Java Persistence Query Language (JPQL), a
platform-independent object-oriented query language which is defined as part of the JPA specifi-
cation.
Example of using Hibernate configuration files:
1 < h i b e r n a t e mapping>
2 < c l a s s name=" Customer " t a b l e =" AppCustomers ">
3 < id name=" id " type =" i n t " column=" id " / >
4 < p r o p e r t y name="name" column="name" type =" S t r i n g " / >
5 < / c l a s s >
6 < / h i b e r n a t e mapping>
Example of using JPA annotations:
1 @Entity
2 @Table ( name=" AppCustomers " )
3 public c l a s s Customer {
4 @Id
5 private i n t id ;
6 S t r i n g name ;
7 [ . . . ]
8 }
Example of using Hibernate Query Language (HQL):
1 public List <Customer > findAllCustomers ( ) {
2 S t r i n g hql = " s e l e c t c from Customer c " ;
3 return executeQuery ( hql ) ;
4 }
5 public L i s t executeQuery ( S t r i n g hql ) {
6 return s e s s i o n . createQuery ( hql ) . l i s t ( ) ;
7 }
8 public Customer findCustomer ( I n t e g e r id ) {
9 return ( Customer ) s e s s i o n . get ( Customer . class , id ) ;
10 }
Focus
• Java RDBA technologies

6. EOSESE – Lille, France, 3 December 2015
Focus
• ORMs promise to reduce the
“object-rela&onal impedance mismatch”
• But do they really?
• What are the risks and advantages of using a par&cular
technology?
• How do projects using ORM evolve over &me?
• How are mul&ple ORM technologies used together? Do
they tend to complement or subs&tute one another?
• How are the code and DB schema kept synchronised?

7. EOSESE – Lille, France, 3 December 2015
Coarse-grained analysis
• Empirical study on open source Java projects
– 13,307 s&ll ac&ve Java projects extracted from GitHub Java
Corpus [Allamanis&Su?on — MSR 2013]
– 4,503 of them used an RDBA technology
– 2,818 projects used JDBC, JPA, Spring or Hibernate

8. EOSESE – Lille, France, 3 December 2015
Coarse-grained analysis
• Evolu&on of RDBA technology usage
– Rela&ve share of each of the 4 considered technologies
• JDBC remains most prominent, but decreasing share since 2006
• Since its introduc&on, share of JPA con&nues to increase

9. EOSESE – Lille, France, 3 December 2015
Coarse-grained analysis
• Co-occurrence of RDBA technologies
– Many projects with JPA, Hibernate or Spring also use
direct JDBC access
– 38% of all projects use at least 2 technologies
– Projects using mul&ple technologies tend to use them
simultaneously (>80%)
JPA
Spring JDBC
Hibernate
284
167 1239
60
86
124
133
139
16
76
61
64143
46
180
Hibernate o8en co-occurs
with JPA (77%) and JDBC (61%)

10. EOSESE – Lille, France, 3 December 2015
Coarse-grained analysis
• Introducing and removal of RDBA technologies
– Technologies are introduced early in project life&me
– Technologies disappear faster in small than in large
projects
Introduc&on &me Removal &me

11. EOSESE – Lille, France, 3 December 2015
Coarse-grained analysis
• Introducing and removal of RDBA technologies
– New technos are introduced more o6en in large projects
– JDBC is not followed frequently by other technos
– Hibernate is frequently followed by another techno
“Survival curve”
Probability that a techno
remains the lastly
introduced one
small projects
large projects

12. EOSESE – Lille, France, 3 December 2015
Fine-grained analysis
• Research ques&ons
• How do developers decide to introduce new RDBA
technologies?
• How and where are par&cular technologies used in the
source code?
• Are database schema elements accessed by different
technologies?
• Focus on 3 par&cular Java projects

13. EOSESE – Lille, France, 3 December 2015
Fine-grained analysis
Distribu&on of accessed tables across technologies
• Hibernate
complements JDBC
• JPA replaces
Hibernate & JDBC
• Tables accessed by
several technologies

14. EOSESE – Lille, France, 3 December 2015
Fine-grained analysis
• Majority of tables
accessed by Hibernate
• A few tables accessed
via JDBC
• Some tables accessed by
both Hibernate &JDBC
Distribu&on of accessed tables across technologies

15. EOSESE – Lille, France, 3 December 2015
Fine-grained analysis
• JPA only used
technology
• High number of
unaccessed tables
Distribu&on of accessed tables across technologies

16. EOSESE – Lille, France, 3 December 2015
Fine-grained analysis
UnIl introducIon of JPA:
majority of files accessing
DB via Hibernate.
A8er introducIon of JPA:
progressive decrease of
#files accessing DB
via Hibernate and JDBC
Distribu&on of files accessing the database

17. EOSESE – Lille, France, 3 December 2015
Fine-grained analysis
Hibernate remains main
technology
JDBC and Hibernate
together in some files
è bad coding prac&ces?
Distribu&on of files accessing the database

18. EOSESE – Lille, France, 3 December 2015
Fine-grained analysis
Stabiliza&on in 2013
Distribu&on of files accessing the database

19. EOSESE – Lille, France, 3 December 2015
Fine-grained analysis
• Majority of Hibernate
mappings replaced by
JPA mappings.
• Big part of DB schema
remains unmapped.
• Small set of tables
contain both
Hibernate and JPA
mappings.
Distribu&on of mapped tables across technologies

20. EOSESE – Lille, France, 3 December 2015
Fine-grained analysis
Hibernate main
technology
è majority of tables
with Hibernate mapping
Distribu&on of mapped tables across technologies

21. EOSESE – Lille, France, 3 December 2015
Fine-grained analysis
JPA only used
technology
è majority of tables
with JPA mapping
Distribu&on of mapped tables across technologies

22. EOSESE – Lille, France, 3 December 2015
• Coarse-grained analysis
– Technologies tend to be used in combina&on (especially for large projects)
– Some technologies (Hibernate) tend to get replaced more o6en
– JPA con&nues to increase its market share
– In spite of ORM promises, “legacy” technologies (JDBC) remain heavily used
– Newly introduced technologies tend to “complement” exis&ng ones
• Fine-grained analysis
– Different projects with very different use of technologies
– Some tables can be accessed through different technologies
– Some source code file used mul&ple technologies
– Some tables/columns are not accessed è might suggest that co-evolu&on
of schema and programs is not always a trivial process
22
Conclusions

23. EOSESE – Lille, France, 3 December 2015
• Broaden the scope
– Non-rela&onal database technologies
– Closed source projects
– Non-Java projects
• Study social aspects
– Different teams for different technologies?
Different teams for code versus database?
• Provide automated tool support
– Decisional support to project managers to choose a new techno
and to decide if, when and how to introduce it
– Visualisa&on support for the code/database co-evolu&on and the
technology co-occurrences
23
Future Work

24. EOSESE – Lille, France, 3 December 2015
• L Meurice, M Goeminne, T Mens, C Nagy, A Decan, A Cleve.
Analyzing the evoluIon of database usage in data-intensive so8ware
systems. [To appear in book chapter]
• M Goeminne, T Mens. Towards a survival analysis of database
framework usage in Java projects. ICSME 2015 ERA track
• M Goeminne, A Decan, T Mens. Co-evolving code-related and
database-related changes in a data-intensive so8ware system.
CSMR-WCRE 2014 ERA track
• L Meurice, A Cleve. DAHLIA: A visual analyzer of database schema
evoluIon. CSMR-WCRE 2014 Tool Demo
• A Cleve, M Gobert, L Meurice, J Maes, J Weber. Understanding
database schema evoluIon: A case study. Science of Computer
Programming (2013)
• A Cleve, T Mens, J-L Hainaut. Data-intensive system evoluIon. IEEE
Computer 43(8): 110-112 (2010)
24
References