Tap Workflows

Slides from a presentation to the Derailed group in Denver about building workflows using Tap (03/2009)

Transcript

1. Tap Workflows a b1 b c1 c2 c ? x x1 x2 y y1 z
2. Tap (Task Application) Framework for command line apps Configurations, Documentation, Distribution Workflows: sequence, fork, merge, sync merge, switch
3. Workflows a --: b --: c Think of a, b, c as methods (literally they are)
4. Sequence a --: b --: c result_a = a(input) result_b = b(result_a) result_c = c(result_b)
5. Fork --a[b, c] result_a = a(input) result_b = b(result_a) result_c = c(result_a)
6. Merge --c{a,b} result_a = a(input) result_b = b(input) result_c = c(result_a, result_b)
7. Variations a --:i b a(input).each do |result_a| # i<terate> b(result_a) end a --:s b b(*a(input)) # s<plat>
8. Variations a --:i b --:a c results_b = [] a(input).each do |result_a| # i<terate> results_b << b(result_a) # a<ggregate> end c(*results_b)
9. Tap Workflows Build workflows from command line Get behavior like in a script
10. Think about | Pipe is just a sequence Serves us well... but it’s just a sequence % git diff | mate
11. Example Digest: splits a protein sequence Fragment: numerical transform on substrings Format Mfg: formats results as .mgf
12. Desired Workflow (protein) - <digest> - [peptides...] (peptide) - <fragment> - [data, headers] (data, headers) - <mgf> - string digest --:i fragment --:s mgf
13. Fragment
14. Fragment
15. Fragment
16. Demo % tap run -- fragment --help % rap fragment --help % rap fragment PEPTIDE % rap fragment PEPTIDE --: dump % rap digest PROTEIN --: dump % rap digest PROTEIN --:i fragment --: dump % rap digest PROTEIN --:i fragment --:a dump % rap digest PROTEIN --:i fragment --:s format_mgf --:a dump
17. Demo % rap digest PROTEIN --:i fragment -- format_mgf -- format_mgf --mz-precision 2 -- dump --no-audit --no-yaml --1[2,3]s “--4{2,3}a” <workflow.yml> - [digest, PROTEIN] - [fragment] - [format_mgf] - [format_mgf, --mz-precision, 2] - [dump, --no-audit, --no-yaml] - 0:1i - 1[2,3]s - 4{2,3}a
18. Why is this good? Tasks are distributable as gems Encapsulation/Reuse Structured enough for novel interfaces
19. Tap-Server Alpha alpha alpha Tasks and Workflows in the browser Immediate, easy interface for new users (customers, friends) Simple Distributable Controllers
20. Demo % tap server http:/ /localhost:8080/ http:/ /localhost:8080/schema % tap run <schema> http:/ /localhost:8080/capture/tutorial % rap load <request.yml> --: tap-http:submit
21. Demo Summary All happens in tap-server Install a gem, get local controllers... tutorials, gui, etc. Potential for interaction with web data Potential to use browser as task interface
22. Goodies... Rap - rakish syntax for declaring tasks Lazydoc - pull comments into object space Configurable - class-based configs mapping to command line Rap.task :sample, :one, :key => ‘value’ do |task, args| task.key # => ‘value’ args.one # => 1 (% rap sample 1) end
23. Gotchas Discovering tasks Task documentation must have ‘::manifest’ tap.yml - used in pwd to manually specify which gems to include; package with gem to allow autodiscovery of gem tasks Rap tasks don’t package, pass results, or appear in tap-server. Local tasks only.
24. tap.rubyforge.org