This page describes the October 1, 2008 revisions to the San programming
language specification.
This is a fairly substantial revision that makes changes to almost
every section, including reordering and renumbering sections. There
are further substantial revisions that are planned. These include
sbutstantially revising section 2 to include language objectives and
a discussion of implementation modes. The sections on sequents and
configuration segments need a substantial rework.
Keywords start with an initial upper case Original text: In the original text identifiers and keywords
were entirely lower case.
New text:Identifiers are still entirely lower case. Keywords are lower
case with the initial letter capitalized. Backslash is no longer a special
character.
Rationale: This change is part of a general program to reduce the
complexity of the language syntax. Keywords and identifiers are now in separate
lexical spaces; there is no potential present or future conflict between keywords
and identifiers.
Impact: Code examples and definitions were changed in almost every
section.
Sections renumbered and reordered Original text: In the original text the order of the major sections
reflected the order in which they written.
New text:Sections were reordered and renumbered.
Rationale: The new order is supposed to provide a more logical
progression.
Impact: Links had to be changed.
Flow control section rewritten Original text: The original section was sort of a grab bag of
flow control constructs.
New text:Constructs were eliminated, e.g., the “filter” constuct,
until loops, compound loops, and the cycle keyword. Two new constructs
were added, Andif and Orif. The escape keyword was replaced by Exitif with
slightly different semantics. The concept of a conditional code chain was
introduced. The switch construct was simplified.
Rationale: The original constructs were somewhat of a mess. In
particular the filter construct was awkward; the Andif and Orif constructs
provide a simpler, cleaner alternative.
Impact: Links had to be changed.
Return verbs were expanded Original text:There was one return verb, Return.
New text:Two new verbs, Fail and Yield, have been added.
Rationale: Two new features have been added. The first
distinguishes between success (Return) and failure (Fail) of an
invocation. The second enables generators and coroutines.
Impact: The section on procs has been revised.
Change print and println to puts and print Original text:Print printed text without a terminating newline;
println printed text with a terminating newline.
New text:Print is replaced by Puts, and println by Print.
Rationale: This is the second rename for these two functions. Originally
they were rawprint and print, which seemed a bit cumbersome. They were changed
to print and println; it was deemed odd to have the more common form longer.
They have been changed to Puts and Print; puts is the C “put string” function.
Impact: Code examples were changed.
The obligatory begin was removed Original text:Block initiators of all kinds had an initial begin
keyword.
New text:The obligatory begin has been removed. The Begin keyword is
still used for simple blocks.
Rationale: The original begin/end formulation was like parentheses in
lisp. The delimiter-keyword-parameters-EOL-body-delimiter structure has been retained but the
keyword also serves as the initial delimiter.
Impact: Code examples were changed.
Appendices were rewritten Original text:There were several appendices covering sundry
categories of keywords. There was no complete list of keywords.
New text:The several appendices were replaced by a single appendix
that is a comprehensive compendium of keywords.
Rationale: The move to initial capital keywords has removed the need for
special cases, so the appendices had to rewritten anyway. Adding a complete list
was desirable.
Impact: Old appendices were replaced and sundry links had to be revised.
Delete naked identifiers Original text: The original specification permitted naked identifiers, e.g. an identifier
beginning with an apostrophe or period that referred to the object the code was in.
It also has the Self keyword to stand for the current object.
New text:Naked identifiers have been eliminated.
Rationale: Naked identifiers came first; it turned out that a Self keyword
was unavoidable. There is no need for having both.
Argument/value pair argument lists have been deleted Original text: The original specification permitted permitted argument/value
argument lists, e.g. if foo is a proc with argument x we could write
foo x=5
New text:Argument/value pair argument lists are no longer permitted.
Rationale: The syntax was ambiguous.
Exponentiation operator changed Original text: The caret, ^, was the logical xor operator and ** was
the exponentiation operator.
New text:The caret is now the exponentiation operator, and there is
no logical xor operator.
Rationale: ** was an anomaly among special characters. Logical xor
is not used often enough to warrant reserving a special character for it.
Executive segment replaced by program Original text: There was an executive segment that essentially listed
a program and a configuration.
New text:The segment has been renamed as Program and the “program”
keyword has been renamed as Entry.
Rationale: Terminology better matches intended usage.
Three levels of agents Original text: A thread could have several principal agents.
New text:There are three levels of agents, a supervisory agent
created by the San execution engine, one or more principal agents
directly below the master agent, and subordinate agents under the principal
agents.
Rationale: This change facilitates having a variety of
execution modes. Thus San might be run as an interactive shell, as
a script file, or as a suite of embedded threads in another program. The
entire notion of threads as a container of agents is being eliminated; agents
are threads and threads are agents.