XConf 2022 - Code As Data: How data insights on legacy codebases can fill the knowledge gap in complex modernization projects.
0 likes52 views
The document explores the challenges of legacy software modernization and how using 'code as data' (cad) can help bridge knowledge gaps by providing insights into legacy codebases. It emphasizes the importance of understanding both known and unknown factors through interactive workshops, tools, and methodologies to create a comprehensive modernization strategy. By transforming programming artifacts into accessible data formats, stakeholders can gain a clearer view of business rules, risks, and system interdependencies.
![© 2022 Thoughtworks | Confidential 7
Java
class HelloWorld {
public static void main(String[]
args) {
System.out.println("Hello, world!");
}
}
C#
using System;
class Program
{
public static void Main(string[]
args)
{
Console.WriteLine("Hello,
world!");
}
}
Python
print("Hello, world!")
Ruby
puts "Hello,
world!"
Scala
object HelloWorld extends
App {
println("Hello, world!")
}
ASP.NET
Response.Write("Hello World!");
Lisp
(princ "Hello, world!")
Haskell
main = putStrLn "Hello, world!"
Malbolge
('&%:9]!~}|z2Vxwv-,POqponl$Hjig%eB@@>}=<M:9wv6WsU
2T|nm-,jcL(I&%$#"
`CB]V?Tx<uVtT`Rpo3NlF.Jh++FdbCBA@?]!~|4XzyTT43Qsq
q(Lnmkj"Fhg${z@>
Hello, (all) World
there is a parser for every programming language…
abb, abnf, acme, agc, alef, fix, algol60, alloy, alpaca, angelscript, space, antlr, apex, apt, argus, arithmetic, asl, asm, asn, aspectj, atl, b, basic, bcl, bcl, bcpl,
bdf, bibcode, bnf, brainflak, brainfuck, c, calculator, callable, capnproto, cayenne, symbol conflicts, clf, clif, clojure, clu, cmake, cobol85, cookie, cool, cpp,
cql, cql3, creole, csharp, css3, csv, ctl, cto, dart2, databank, dcm, dgol, dice, dif, doiurl, dot, edif300, edn, erlang, fasta, fdo91, fen, flatbuffers, flowmatic,
fixes, focal, fol, fortran77, fusion-tables, gdscript, gedcom, gff3, gml, golang, graphql, graphstream-dgs, gtin, guido, guitartab, haskell, html, http, hypertalk,
icalendar, icon, idl, inf, informix, infosapient, iri, iso8601, istc, itn, first commit, janus, java, javadoc, javascript, joss, jpa, json, json5, karel, karel, kirikiri-tjs,
kotlin, kquery, kuka, lambda, lark, lcc, less, limbo, lisa, logo, lolcode, loop, lpc, lrc, ltl, lua, lucene, matlab, mckeeman-form, mdx, memcached_protocol,
metamath, metric, microc, modelica, modula2pim4, molecule, moo, morsecode, mps, muddb, mumath, mumps, muparser, nanofuck, newick, oberon, objc,
oncrpc, orwell, p, parkingsign, pascal, pcre, pddl, pdn, peoplecode, pgn, php, pii, pike, pl0, plucid, fix, ply, pmmn, postalcode, powerbuilder, powerbuilderdw,
powerquery, prolog, promql, propcalc, properties, protobuf2, protobuf3, prov-n, python, qif, quakemap, r, racket-bsl, racket-isl, rcs, redcode, refal,
comments, rego, restructuredtext, rexx, rfc1035, rfc1960, rfc3080, rfc822, robotwars, romannumerals, rpn, ruby, rust, last month, scala, scotty, scss, last
month, sexpression, sgf, sharc, sici, sickbay, sieve, smalltalk, smiles, smtlibv2, snobol, snowball, more, solidity, sparql, spass, sql, stacktrace, stellaris, stl,
stringtemplate, suokif, swift-fin, swift, tcpheader, teal, telephone, terraform, thrift, tiny, tinybasic, tinyc, tinymud, tinyos_nesc, tl, tnsnames, tnt, toml, trac,
tsv, ttm, turing, turtle-doc, turtle, unicode, unreal_angelscript, upnp, url, useragent, v, vb6, vba, velocity, verilog, vhdl, vmf, wat, wavefront, webidl, wkt, wln,
wren, xml, xpath, xsd-regex, xyz, z
opensource ANTLR grammars available at https://github.com/antlr/grammars-v4
…or you can use](https://image.slidesharecdn.com/xconf2022-codeasdataalessandroconfetti-220708082726-2e1a1d1e/85/XConf-2022-Code-As-Data-How-data-insights-on-legacy-codebases-can-fill-the-knowledge-gap-in-complex-modernization-projects-7-320.jpg)
![© 2022 Thoughtworks | Confidential
“Challenges in data
visualization does not actually
involve visualizing Data. [...] The
challenge is in crafting a
visualization that is easily
reusable, composable, and
extensible.”
34
A. Bergel, Agile Visualization
Lulu Press, 2016, ISBN 978136531409
34
© 2022 Thoughtworks](https://image.slidesharecdn.com/xconf2022-codeasdataalessandroconfetti-220708082726-2e1a1d1e/85/XConf-2022-Code-As-Data-How-data-insights-on-legacy-codebases-can-fill-the-knowledge-gap-in-complex-modernization-projects-34-320.jpg)
![© 2022 Thoughtworks | Confidential
39
What is a
meta-model?
When we need to explore and reason
about complex systems, we need to find
the right kind of representation
(i.e. the right questions we need answer to).
A pragmatic way to find the right balance
between accuracy and outcome.
Concrete
Abstract Meta-[...]-Model
describes
e.g. alphabet, numbers, units,
colors, cartographic projection
Symbols and grammar to represent structure
and vocabulary of a valid meta-model.
Meta-Model
describes
e.g. map legend and conventions
Structure and vocabulary of a valid model.
Model
represents
e.g. street map
Simplified representation of the problem, driven
by questions we need answered.
Subject/Problem
e.g. route between two cities
Something we want to reason about Complex
Simplified
See for reference: J. Bezivin and O. Gerbe, Towards a precise definition of the OMG/MDA
framework,
Proceedings 16th AICASE (ASE 2001), 2001, pp. 273-280, doi: 10.1109/ASE.2001.989813.
If we oversimplify it, we may end up with lot
of underestimated or unmitigated risks.
If we overcomplicate it, we may easily enter
never-ending rabbit-holes and struggle to
deliver the overall picture.](https://image.slidesharecdn.com/xconf2022-codeasdataalessandroconfetti-220708082726-2e1a1d1e/85/XConf-2022-Code-As-Data-How-data-insights-on-legacy-codebases-can-fill-the-knowledge-gap-in-complex-modernization-projects-39-320.jpg)
XConf 2022 - Code As Data: How data insights on legacy codebases can fill the knowledge gap in complex modernization projects.
- 1. © 2022 Thoughtworks | Confidential CaD: (Code As Data) How data insights on legacy codebases can fill the knowledge gap in complex modernization projects
- 2. © 2022 Thoughtworks | Confidential Why legacy modernization is always so challenging? 2 We usually rely on people who built the software (i.e. DEVs) or the ones dealing with it (e.g. SMEs, Users) to collect knowledge about it (e.g. what it does?, why?). But people come and go and knowledge is scattered or lost… What if source code could tell some interesting facts about itself too? 1998 2016
- 3. © 2022 Thoughtworks | Confidential 3 What People Think Code Is CaG C(ode) a(s) G(ibberish) What Developers Think Code Is CaL C(ode) a(s) L(iterature) What Computers Think Code Is CaD C(ode) a(s) D(data)
- 4. © 2022 Thoughtworks | Confidential 4 #include <stdio.h> main( ) { printf("hello, world"); } Hello, World Kernighan, Brian W.; Ritchie, Dennis M. (1978). The C Programming Language (1st ed.). Englewood Cliffs, NJ: Prentice Hall. ISBN 0-13-110163-3. (developer view)
- 5. © 2022 Thoughtworks | Confidential $ clang -Xclang -ast-dump -fsyntax-only hello-world.c 1 5 10 15 20 25 30 1 # i n c l u d e < s t d i o . h > 2 3 m a i n ( ) { 4 p r i n t f ( “ h e l l o , w o r l d “ ) ; 5 } 5 where token who what semantics Hello, World (computer view) AST syntax Abstract Syntax Tree ↳ ↳ ↳ ↳ ↳ ↳ ↳
- 6. © 2022 Thoughtworks | Confidential 6 TYPE ID PARENT_ID FILE LINE COLUMN TEXT FunctionDecl 0x7fedd489f830 hello-world.c 3 main CompoundStmt 0x7fedd489fa10 0x7fedd489f830 hello-world.c 3 12 CallExpr 0x7fedd489f9b8 0x7fedd489fa10 hello-world.c 4 3 ImplicitCastExpr 0x7fedd489f9a0 0x7fedd489f9b8 hello-world.c 4 3 DeclRefExpr 0x7fedd489f8d0 0x7fedd489f9a0 hello-world.c 4 3 printf ImplicitCastExpr 0x7fedd489f9f8 0x7fedd489f9b8 hello-world.c 4 10 ImplicitCastExpr 0x7fedd489f9e0 0x7fedd489f9f8 hello-world.c 4 10 StringLiteral 0x7fedd489f928 0x7fedd489f9e0 hello-world.c 4 10 Hello, World!n $ clang -Xclang -ast-dump -fsyntax-only hello-world.c 1 5 10 15 20 25 30 1 # i n c l u d e < s t d i o . h > 2 3 m a i n ( ) { 4 p r i n t f ( “ h e l l o , w o r l d “ ) ; 5 } where token who what semantics Hello, World (data view) AST syntax Abstract Syntax Tree ↳ ↳ ↳ ↳ ↳ ↳ ↳
- 7. © 2022 Thoughtworks | Confidential 7 Java class HelloWorld { public static void main(String[] args) { System.out.println("Hello, world!"); } } C# using System; class Program { public static void Main(string[] args) { Console.WriteLine("Hello, world!"); } } Python print("Hello, world!") Ruby puts "Hello, world!" Scala object HelloWorld extends App { println("Hello, world!") } ASP.NET Response.Write("Hello World!"); Lisp (princ "Hello, world!") Haskell main = putStrLn "Hello, world!" Malbolge ('&%:9]!~}|z2Vxwv-,POqponl$Hjig%eB@@>}=<M:9wv6WsU 2T|nm-,jcL(I&%$#" `CB]V?Tx<uVtT`Rpo3NlF.Jh++FdbCBA@?]!~|4XzyTT43Qsq q(Lnmkj"Fhg${z@> Hello, (all) World there is a parser for every programming language… abb, abnf, acme, agc, alef, fix, algol60, alloy, alpaca, angelscript, space, antlr, apex, apt, argus, arithmetic, asl, asm, asn, aspectj, atl, b, basic, bcl, bcl, bcpl, bdf, bibcode, bnf, brainflak, brainfuck, c, calculator, callable, capnproto, cayenne, symbol conflicts, clf, clif, clojure, clu, cmake, cobol85, cookie, cool, cpp, cql, cql3, creole, csharp, css3, csv, ctl, cto, dart2, databank, dcm, dgol, dice, dif, doiurl, dot, edif300, edn, erlang, fasta, fdo91, fen, flatbuffers, flowmatic, fixes, focal, fol, fortran77, fusion-tables, gdscript, gedcom, gff3, gml, golang, graphql, graphstream-dgs, gtin, guido, guitartab, haskell, html, http, hypertalk, icalendar, icon, idl, inf, informix, infosapient, iri, iso8601, istc, itn, first commit, janus, java, javadoc, javascript, joss, jpa, json, json5, karel, karel, kirikiri-tjs, kotlin, kquery, kuka, lambda, lark, lcc, less, limbo, lisa, logo, lolcode, loop, lpc, lrc, ltl, lua, lucene, matlab, mckeeman-form, mdx, memcached_protocol, metamath, metric, microc, modelica, modula2pim4, molecule, moo, morsecode, mps, muddb, mumath, mumps, muparser, nanofuck, newick, oberon, objc, oncrpc, orwell, p, parkingsign, pascal, pcre, pddl, pdn, peoplecode, pgn, php, pii, pike, pl0, plucid, fix, ply, pmmn, postalcode, powerbuilder, powerbuilderdw, powerquery, prolog, promql, propcalc, properties, protobuf2, protobuf3, prov-n, python, qif, quakemap, r, racket-bsl, racket-isl, rcs, redcode, refal, comments, rego, restructuredtext, rexx, rfc1035, rfc1960, rfc3080, rfc822, robotwars, romannumerals, rpn, ruby, rust, last month, scala, scotty, scss, last month, sexpression, sgf, sharc, sici, sickbay, sieve, smalltalk, smiles, smtlibv2, snobol, snowball, more, solidity, sparql, spass, sql, stacktrace, stellaris, stl, stringtemplate, suokif, swift-fin, swift, tcpheader, teal, telephone, terraform, thrift, tiny, tinybasic, tinyc, tinymud, tinyos_nesc, tl, tnsnames, tnt, toml, trac, tsv, ttm, turing, turtle-doc, turtle, unicode, unreal_angelscript, upnp, url, useragent, v, vb6, vba, velocity, verilog, vhdl, vmf, wat, wavefront, webidl, wkt, wln, wren, xml, xpath, xsd-regex, xyz, z opensource ANTLR grammars available at https://github.com/antlr/grammars-v4 …or you can use
- 8. © 2022 Thoughtworks | Confidential Compilers use all metadata to translate code into executable instructions. What is all this for? Is there anything that can be leveraged by Business Analysts, Project Managers, or IT managers at large*? 8 Static code analysis tools (e.g. linters) use AST metadata to identify potential issues (e.g. programming errors, bugs, non idiomatic code, and suspicious constructs, metrics). IDEs use filtered metadata (e.g. variables, functions, classes, methods) to provide navigation, hints, and code completion. DEVs use actually the same metadata (unconsciously) to read the code! * and DEVs too…
- 9. © 2022 Thoughtworks | Confidential 9 Known Knowns Identified Knowledge Known Unknowns Identified Risk Unknown Knowns Untapped Knowledge Unknown Unknowns Unidentified Risk Legacy Modernization Challenges… proactively reactively discoveries, inceptions, user stories, acceptance criteria spikes, RAIDs, modernization patterns cross-functional teams, short iterations, IPMs … where Agile helps Agile tools & techniques help to proactively address KK & KU and reactively UK & UU.
- 10. © 2022 Thoughtworks | Confidential CaD can help to proactively mitigate the risks about UK & UU. … where can CaD help? 10 Known Knowns Identified Knowledge Known Unknowns Identified Risk Unknown Knowns Untapped Knowledge Unknown Unknowns Unidentified Risk proactively Legacy Modernization Challenges… proactively reactively
- 11. © 2022 Thoughtworks | Confidential Example n. 1 - Unknown Knowns project-level support for BAs & tech analysis 11
- 12. © 2022 Thoughtworks | Confidential Project-level risks mitigation Use Case: Modernization of a Pricing Engine We were asked to replace a pricing engine under development for the past 30 years. 12 We went through an inception and several workshops with stakeholders, SMEs, DEVs. We collected all the available knowledge (KK), and identified all the grey areas that would require further investigations (KU). Are these really just all the business rules? We found out that business rules were encoded as table rows (e.g. exception/inclusion rules) or field values (e.g. operation rules and values), referenced and manipulated inside legacy code. SMEs and DEVs told us that there are only 60 tables to care about…
- 13. © 2022 Thoughtworks | Confidential 13 Inception Proactively untap knowledge Consolidate SMEs knowledge BAs Refine the project scope PO Trigger SMEs conversations BAs Explore grey areas BAs identify missing tables/proc identify referenced fields Explore legacy code Parse legacy code DEVs DEVs BAs BAs CaD Pipeline & Tools Goals: ● Tactical: it should not require a huge investment in time and resources (i.e. DEVs should not have to become legacy code experts) ● Pragmatic: just search for possible clues (e.g. tables, fields, procedures not mentioned in the workshops) ● Accessible: BAs and SMEs should be able to use and explore the outcomes (i.e. use tools they already know) Use Case: Modernization of a pricing engine
- 14. © 2022 Thoughtworks | Confidential CaD Pipeline Goals ● Tactical: ~300 clojure LoC leveraging an existing open-source legacy language ANTLR-based parser. ● Pragmatic: leveraging semantic features of the legacy language to filter tokens (never underestimate the expressiveness of an old programming language) ● Accessible: the output was a spreadsheet that could be easily filtered by table and column name, or explored with pivot tables. All the tokens were connected (via Excel hyperlinks) to tables documentation and specific line/column of the source code in VS Code (with syntax coloring thanks to an open-source plugin). DB Catalog Parse Tables & Columns Metadata Tables/Fields Names 1.1k tables 11k fields Table List Parse Project Scope SMEs 60 tables ??? fields Filter & Merge Tokens referencing Tables & Fields 128 tables 1k fields 4.7k tokens Source Code Parse P r o j e c t S c o p e 4M LoC 22k LoC 117k tokens + Browser Online Docs BAs/SMEs XLSX Excel BAs VS Code 4th gen language VS Code plugin DEVs/SMEs
- 15. © 2022 Thoughtworks | Confidential 15 Easy interoperability with Java libraries Easy access to ANTRL objects and attributes and XLSX libraries to read/write large files. Fast in-memory parallel data transformations Clojure transducers and core.async libraries provide easy & fast parallel in-memory transformations without requiring huge resources or infra. REPL driven development The REPL allow an instant feedback workflow that can dramatically speed up exploring Java libraries and data structures. Why Clojure? Clojure is a fast modern Lisp that runs on top of the JVM (and CRI/V8 too). because we love parenthesis ;)
- 16. © 2022 Thoughtworks | Confidential How it looks like 16 Excel Spreadsheet VS Code editor: token context Token Table’s online docs Hyperlink Hyperlink Dataset Source Code DB SME DEVs/SMEs BAs/SMEs BAs ● parse unit path: source file path ● file path: original source file path (may be different in case of include file) ● line: token line inside the source file ● column: token starting column inside the line ● source docs: link to VS code to highlight the token inside the source file ● type: token semantics tag ● text: token actual text ● node id: token id (parse unit context) ● parent id: token AST parent ● level: token AST indentation ● procedure id: procedure uuid ● procedure name: procedure name ● table name: matching table ● column name: matching column ● table docs: link to table’s online docs ● ambiguous term: true/false ● in scope: true/false
- 17. © 2022 Thoughtworks | Confidential 17 Are these really just all the business rules? Use Case: Modernization of a pricing engine CaD outcomes ● Tables: +40% more tables in scope (some were edge cases, other seldom used) ● Fields: scope reduced to 36% of fields (most fields were used for other purposes) ● Business Rules: whenever there was a computation issue we could go to the exact point in the source code to clarify assumptions and behaviors
- 18. © 2022 Thoughtworks | Confidential Example n. 2 - Unknown Unknowns program-level support for legacy modernization 18
- 19. © 2022 Thoughtworks | Confidential Program-level risks mitigation Use Case: legacy ERP modernization We were asked to replace an existing monolithic on-prem ERP-like system made of several modules, and under development for the past 30 years. 19 We went through a discovery and several workshops with DEVs, OPSs, DBAs, SMEs, and business stakeholders. We defined a target functional & tech architecture (KK), and identified modernisation patterns & RAIDs (KU) with tentative mitigations. how much is going to cost? where should we start from? We found out that the system was integrated with several business processes, exchanging data with many applications, and everybody was scared of breaking something…
- 20. © 2022 Thoughtworks | Confidential 20
- 21. © 2022 Thoughtworks | Confidential 21 There should be a wire somewhere on your left… Or maybe it’s on the right… but don’t cut the other ones!
- 22. © 2022 Thoughtworks | Confidential We collected an amazing amount of information inside stickies of different colors and shape. What if we could translate stickies into data? 22 22 © 2022 Thoughtworks
- 23. © 2022 Thoughtworks | Confidential 23 Discovery & Workshops Proactively uncover unknown risks Consolidate FIndings BAs Refine the program roadmap PO Learn from mistakes BAs Start several Inceptions/deliveries BAs Use Case: legacy ERP modernization Goals: ● Strategical: it should help defining a long term plan backed by data and KPIs that can evolve over time. ● Comprehensive: it should cover the entire applications landscape not just a single project. ● Flexible: it should quickly provide answer to basic questions, but also support further investigations. CaD Pipeline & Tools Convert Stickies to Data DEVs Parse Code & Merge with Stickies Data DEVs Explore & Compute KPIs DEVs DEVs Collect Projects Metrics BAs
- 24. © 2022 Thoughtworks | Confidential ..to data (visualization) Area 3 M3 M2 M1 Area 2 M3 M2 M1 Area 4 M3 M2 M1 Area 1 M3 M2 M1 from stickies… 1. We split the monolith into logical Areas & Modules (both existing and new ones)
- 25. © 2022 Thoughtworks | Confidential ..to data (visualization) Area 3 M3 M2 M1 Area 2 M3 M2 M1 Area 4 M3 M2 M1 Area 1 M3 M2 M1 Tables from stickies… 1. We split the monolith into logical Areas & Modules (both existing and new ones) 2. We map Tables belonging to just a single Module (if any)
- 26. © 2022 Thoughtworks | Confidential ..to data (visualization) Area 3 M3 M2 M1 Area 2 M3 M2 M1 Area 4 M3 M2 M1 Area 1 M3 M2 M1 APIs Tables from stickies… 1. We split the monolith into logical Areas & Modules (both existing and new ones) 2. We map Tables belonging to just a single Module (if any) 3. We map APIs belonging to just a single Module (if any)
- 27. © 2022 Thoughtworks | Confidential ..to data (visualization) Area 3 M3 M2 M1 Area 2 M3 M2 M1 Area 4 M3 M2 M1 Area 1 M3 M2 M1 APIs Tables Source Code from stickies… 1. We split the monolith into logical Areas & Modules (both existing and new ones) 2. We map Tables belonging to just a single Module (if any) 3. We map APIs belonging to just a single Module (if any) 4. We parse source code to identify chain of calls (who calls whom) and access to tables (who reads/writes data where)
- 28. © 2022 Thoughtworks | Confidential ..to data (visualization) Area 3 M3 M2 M1 Area 2 M3 M2 M1 Area 4 M3 M2 M1 Area 1 M3 M2 M1 APIs Tables Source Code from stickies… 1. We split the monolith into logical Areas & Modules (both existing and new ones) 2. We map Tables belonging to just a single Module (if any) 3. We map APIs belonging to just a single Module (if any) 4. We parse source code to identify chain of calls (who calls whom) and access to tables (who reads/writes data where) 5. We map APIs to source code entry points (e.g. functions)
- 29. © 2022 Thoughtworks | Confidential we can now explore each area, module, table, procedure, or API and follow interactively all the trails that connect stickies to source code. 29 29 © 2022 Thoughtworks
- 30. © 2022 Thoughtworks | Confidential CaD Pipeline Goals ● Strategical: integrating workshop outcomes with technical catalogs, we can intersect target and current state (e.g. sizing target features complexity slicing current implementation). ● Comprehensive: the pipeline can be easily extended to include more languages, projects, or artefacts (e.g. configuration files, parsable documentation) ● Flexible: leveraging meta-models we can explore source code in a guided way or build our own way through it and identify risks and areas to deep-dive (e.g. shared dependencies, domain bleeding, domain complexity). Build Annotated Graph Source Code Parse + Discovery Stickies Parse SMEs Tables & APIs Annotations DB Catalog APIs Catalog Parse Pharo DEVs/SMEs Explore Graph & Build KPIs Refine Roadmap BAs/SMEs Excel Merge KPIs & Metrics
- 31. © 2022 Thoughtworks | Confidential 31 Complex data structures visualization tools Pharo integrate Roassal library to display complex interactive graph-oriented data structures. Dynamic graphical inspector Every object can be explored with a graphical inspector and may define custom views based on Roassal. Pause & resume support In every moment, we can pause the exploration and save objects/views to disk and restart later from where we left. Why Pharo? Pharo is a fast modern Smalltalk focused on simplicity and immediate feedback. because we love objects soup ;)
- 32. © 2022 Thoughtworks | Confidential 32 32 Lesson learned so far Do we have answers? Not yet, but we started to collect evidences not just gut feelings. Look for what overlaps (e.g. shared libraries, table accessed by several modules) to anticipate possible issues Look for what matches to collect data about effort (e.g. LoC). Look for what doesn’t match (e.g. table not accessed by code, code not invoked by other code or API) to uncover unseen risks.
- 33. © 2022 Thoughtworks | Confidential “In contrast to visual programming and diagramming for software design, software visualization is not so much concerned with the construction, but with the analysis of programs and their development process.” 33 S. Diehl, Software Visualization Springer, 1998, ISBN 9783540465041 33 © 2022 Thoughtworks
- 34. © 2022 Thoughtworks | Confidential “Challenges in data visualization does not actually involve visualizing Data. [...] The challenge is in crafting a visualization that is easily reusable, composable, and extensible.” 34 A. Bergel, Agile Visualization Lulu Press, 2016, ISBN 978136531409 34 © 2022 Thoughtworks
- 35. © 2022 Thoughtworks | Confidential $ tail -f questions Alessandro Confetti Tech Principal [email protected] 35
- 36. © 2022 Thoughtworks | Confidential 36
- 37. © 2022 Thoughtworks | Confidential 37 W I P
- 38. © 2022 Thoughtworks | Confidential 38 Hello, World main( ) { extern a, b, c; putchar(a); putchar(b); putchar(c); putchar('!*n'); } a 'hell'; b 'o, w'; c 'orld'; Kernighan, Brian W. (1972). A Tutorial Introduction to the Language B. Bell Laboratories (p 4)
- 39. © 2022 Thoughtworks | Confidential 39 What is a meta-model? When we need to explore and reason about complex systems, we need to find the right kind of representation (i.e. the right questions we need answer to). A pragmatic way to find the right balance between accuracy and outcome. Concrete Abstract Meta-[...]-Model describes e.g. alphabet, numbers, units, colors, cartographic projection Symbols and grammar to represent structure and vocabulary of a valid meta-model. Meta-Model describes e.g. map legend and conventions Structure and vocabulary of a valid model. Model represents e.g. street map Simplified representation of the problem, driven by questions we need answered. Subject/Problem e.g. route between two cities Something we want to reason about Complex Simplified See for reference: J. Bezivin and O. Gerbe, Towards a precise definition of the OMG/MDA framework, Proceedings 16th AICASE (ASE 2001), 2001, pp. 273-280, doi: 10.1109/ASE.2001.989813. If we oversimplify it, we may end up with lot of underestimated or unmitigated risks. If we overcomplicate it, we may easily enter never-ending rabbit-holes and struggle to deliver the overall picture.
- 40. © 2022 Thoughtworks | Confidential ● Can be used to describe different kind of common diagrams (e.g. E/R, UML), semantics for hierarchical structures (e.g. XML, JSON), or programming languages (e.g. procedural, functional, object-oriented). 40 What is a meta-model? When we need to explore and reason about complex systems, we need to find the right kind of representation (i.e. the right questions we need answer to). A pragmatic way to find the right balance between accuracy and outcome. If we oversimplify it, we may end up with lot of underestimated or unmitigated risks. If we overcomplicate it, we may easily enter never-ending rabbit-holes and struggle to deliver the overall picture. FAMIX meta-model FAME meta-meta-model Can be used to describe different kind of common diagrams (e.g. E/R, UML), semantics for hierarchical structures (e.g. XML, JSON), or programming languages (e.g. procedural, functional, object-oriented). Support both procedural and object oriented languages. Plugins available for many languages (e.g. C/C++, C#, Clojure, Java, JavaScript, JSX/React, PHP). MSE file format can be used to export/import models based on FAMIX.
- 41. © 2022 Thoughtworks | Confidential FAME meta-meta-model Description family of meta-meta-models for describing and defining meta-models ● All meta-models share a series of common features and basic enquiring capabilities. ● Can be used to describe different kind of common diagrams (e.g. E/R, UML), semantics for hierarchical structures (e.g. XML, JSON), or programming languages (e.g. procedural, functional, object-oriented).
- 42. © 2022 Thoughtworks | Confidential FAMIX meta-model Description family of meta-models for representing the structure of software projects. ● Support both procedural and object oriented languages. ● Plugins available for many languages (e.g. C/C++, C#, Clojure, Java, JavaScript, JSX/React, PHP). ● MSE file format can be used to export/import models based on FAMIX. ● All models share a series of common features and basic enquiring capabilities (e.g. dependency trees).
- 43. © 2022 Thoughtworks | Confidential 43 43 There is an growing community of researchers and tools We are not alone…