todo.txt

- provide help for matrix.svd function
- modify get/set methods for gdt table to accept column's names
- rename gdt.new to gdt.alloc and define gdt.new to initialize the table
  like matrix.new
- in gdt.lineplot remove the parameter name from legend if only one parameter
  is plotted. It can be put in the title instead.
- sanitize variable names in gdt-lm macro generation => exclude Lua keywords and
  headers that cannot be considered as Lua identifiers
- add more colors and markers for generated plots
- add a "correlation" function for gdt table like GNU R
- carry out a complete example for PCA analysis in GNU R book (heptatlon)
- bug with PREDICTED [[th_sigeb.ell] (PREDICTED)]
- introduire a WHERE clause to filter data
- ajouter le Degree of freedom dans le summary du gdt.lm
- DONE/ dans PREDICTED il devrait verifier si la colonne n'existe pas déjà
- dans gdt.plot il faut skipper les lignes undefinis au niveau de champs
  de aggregation
- implement col_insert/append using expressions

A REFLECHIR
- trois mechanisms sont disponible pour travailler sur les tables (gdt)
  - functional approach: t:col_append("boo", |r| r.a + r.b)
  - simple string header like: gdt.plot(t, {"wafer", "site"}, "thickness")
  - GNU R like LM model specifications: "thickness ~ 1, exec_time, tool"