Introduction to analysis description languages

• Sezen Sekmen (Kyungpook National University (KR))

Room: Sunrise (WH11NE)

ADL and the transpiler adl2tnm

• Harry Prosper (Florida State University (US))

Room: Sunrise (WH11NE)

CutLang: analysis description language and runtime interpreter

• Gokhan Unel (University of California Irvine (US))

Room: Sunrise (WH11NE)

ADL/CutLang: hands-on demo

• Sezen Sekmen (Kyungpook National University (KR))
• Gokhan Unel (University of California Irvine (US))
• Harry Prosper (Florida State University (US))

Room: Sunrise (WH11NE)

ADL/CutLang: hands-on demo (continued)

• Sezen Sekmen (Kyungpook National University (KR))
• Gokhan Unel (University of California Irvine (US))
• Harry Prosper (Florida State University (US))

Room: Sunrise (WH11NE)

lhada2rivet

• Philippe Gras (Université Paris-Saclay (FR))

Room: Sunrise (WH11NE)

How to build your own language: hands-on demo

• Jim Pivarski (Princeton University)

Room: Sunrise (WH11NE)

LINQ

• Gordon Watts (University of Washington (US))

Room: Sunrise (WH11NE)

YAML as an ADL

• Benjamin Krikler (University of Bristol (GB))

Room: Sunrise (WH11NE)

NAIL: A prototype analysis language on top of RDataFrame

• Andrea Rizzi (INFN Sezione di Pisa, Universita' e Scuola Normale Superiore, P)

Room: Sunrise (WH11NE)

TTreeFormula

• Philippe Canal (Fermi National Accelerator Lab. (US))

Room: Sunrise (WH11NE)

AEACUS and RHADAMANTUS

• Joel Walker (Sam Houston State University)

Room: Sunrise (WH11NE)

Discussion: where/for what do we need a domain-specific language?

• Harry Prosper (Florida State University (US))

Room: Sunrise (WH11NE) * What different purposes can ADLs serve? * One or multiple ADLs? * The scope of ADLs? selecting events? filling histograms? fitting? * How would ADLs improve our way of thinking about our analyses? * ...

Discussion: what physics-specific content should be included?

• Sezen Sekmen (Kyungpook National University (KR))

Room: Sunrise (WH11NE) * Which elements of object and event operations can be described in an ADL? Where do we stop? * What logical / mathematical operations do we use? * How far do we go in describing composite objects, object groups, event variables? * ...

Discussion: what kind of language/syntax do we need?

• Jim Pivarski (Princeton University)

Room: Sunrise (WH11NE) * Most generic and economic ways to cover the widest collection of needs from the biggest variety of analyses? * Objects, composite objects, event variables, definitions, selections, event weighting, … * Intrinsic loops, reducers, optimizers, object combinations/permutations/partitioning, ternary operations, … * Expressing the event input * Expressing external functions * ...

Discussion: what tooling (parser/interpreter/compiler) do we need?

• Jim Pivarski (Princeton University)

Room: Sunrise (WH11NE) * Interpreters? Transpilers? Which method for which purpose? * How to create the most flexible tools for most generic usage? Issues of speed, user-friendliness? * How to automate incorporation of event input formats? How to deal with different event formats? Common extensible object(s)? * How to automate incorporation of external functions? * Histogramming, event weighting? * ...

Discussion: ADLs for analysis preservation

• Pamfilos Fokianos (CERN)
• Sunje Dallmeier-Tiessen (Humboldt-Universitaet zu Berlin (DE))
• Tibor Simko (CERN)

Room: Sunrise (WH11NE) * What are the central LHC analysis preservation tools? * How can ADLs help analysis preservation? How can they be incorporated into the system? * Databases for event input formats, standard object definitions, event variables, external functions… ? * ...

Discussion: Feedback from other experiments

• Tom Junk (Fermi National Accelerator Lab. (US))

Room: Sunrise (WH11NE) * What do physics analyses consist of in other experiments? * Would ADLs for other experiments be built on similar concepts as in ATLAS/CMS? * What can ATLAS/CMS learn from the other experiments? * ...

Summary / How to move forward

• Sezen Sekmen (Kyungpook National University (KR))
• Stephen Mrenna (FERMILAB)
• Alexx Perloff (University of Colorado Boulder (US))
• Gokhan Unel (University of California Irvine (US))
• Jim Pivarski (Princeton University)
• Harry Prosper (Florida State University (US))

Room: Sunrise (WH11NE) Summarizing the next steps and planning how to proceed.

Fermilab colloquium : Particle physics and programming languages

• Jim Pivarski (Princeton University)

Room: Wilson Hall Programming languages aren't for computers; they're for people. If a language doesn't make it easier to express your physics problem, it's not a suitable language. Some fields have specialized "Domain Specific Languages" (DSLs) that trade freedom of expression for focus on the problem at hand, and can even improve performance by limiting this scope. A prime example is SQL, widely used by data analysts outside of physics, which trades generic computation for a SELECT-WHERE-GROUPBY pattern. Interestingly, this was the design pattern of the first electromechanical computers (Hollerith machine, 1890) and it's still a major focus of big data today (Dean & Ghemawat: MapReduce, 2004). Particle physics problems don't fit SQL well; in fact, physicists became involved in computing in tandem with the invention of generic, digital computers (Von Neumann's stored-program machine, 1945). I will present some history, some general features of programming languages, what "declarative" really means, and will show some perhaps surprising examples of DSLs you're already using.