lib/sh/make.sh: remaking alias: documented that you MUST use set_make_vars at the beginning of any function that uses this, so that $_remake is properly set to $remake and not left at its previous value

moved everything into /trunk/ to create the standard svn layout, for use with tools that require this (eg. git-svn). IMPORTANT: do NOT do an `svn up`. instead, re-use your working copy's existing files with `svn switch` (http://svnbook.red-bean.com/en/1.6/svn.ref.svn.c.switch.html).

fix: lib/sh/util.sh: verbosity_min(): usage: clarified that '' is a special value that causes $verbosity to be overwritten to ''

lib/sh/local.sh: public_schema_exists(): use a higher log_level for pg_schema_exists, to avoid all the verbose output involved in running the query

bugfix: lib/sh/local.sh: public_schema_exists(): can no longer use psql_script_vegbien for this, because using `SET search_path` (called by psql_script_vegbien) with a schema that does not exist no longer produces an error. instead, use new pg_schema_exists(), which uses a different command that does produce an error if the schema does not exist.

lib/sh/db.sh: added pg_require_schema()

lib/sh/util.sh: stderr2stdout(): documented that this redirects fd 2->1 and log_fd (but not back to 2)

bugfix: lib/sh/util.sh: stderr2stdout() use `command` before tee, which re-filters log_fd so that stderr itself is also filtered. this allows log-filtering out an otherwise-confusing benign error when using e.g. stderr_matches().

lib/sh/util.sh: added not(), for use in prefixing wrapped commands

lib/sh/db.sh: added pg_schema_exists()

lib/sh/util.sh: added stderr_matches()

lib/sh/util.sh: documented that fds 2x/3x should not be used because we use these, as opposed to 1x which is used by the shell internally

lib/sh/util.sh: added stdout_contains()

lib/sh/util.sh: added stderr2stdout()

fix: lib/sh/db.sh: pg_table_exists(): usage: documented that $table is actually required for this function

lib/sh/util.sh: import_vars: don't overwrite vars that are already defined, to allow the caller to specify their own values for the vars to create. this requires callers that rely on the overwriting functionality to reverse the order in which they run use_* commands, so that the higher-precedence use_* is applied first and the other one as the default values for the first.

lib/sh/db.sh: pg_table_exists(): use `SELECT NULL` instead of `SELECT *` to avoid a long column list cluttering up the log output

lib/sh/db.sh: added pg_table_exists()

lib/sh/util.sh: already_exists_msg(): added instructions on how to force-remake when the file already exists (prepend `rm=1` to the command)

bugfix: lib/sh/make.sh: $remake: need to explicitly propagate this to invoked commands if it was set from $rm

lib/sh/db.sh: mk_select(): usage: documented that this also takes a $limit/$n param

lib/sh/db.sh: limit(): also support using $n as the limit param, since this var name is used by other parts of the import process

lib/sh/db.sh: limit(): usage: documented that this also need a $limit param

lib/sh/local.sh: psql(): also accept $public as the $schema param, since this is used by a lot of import scripts

lib/sh/util.sh: added require_dot_script()

bugfix: lib/sh/util.sh: $top_script: use @BASH_SOURCE instead of $0, because this is also defined for .-scripts

bugfix: lib/sh/util.sh: is_dot_script(): need to subtract 1 from ${#BASH_LINENO[@]}, because this is the array length rather than the index of the last element as in Perl

lib/sh/util.sh: added is_dot_script()

bugfix: lib/sh/util.sh: alias_append(): need to enclose $(alias) call in "" because its result may contain separator chars (i.e. whitespace) that will be parsed incorrectly. this appears to only be a bug when runscripts are run as shell-includes, with a leading ".".

bugfix: lib/sh/local.sh: psql(): $is_root: use `` around case statement instead of $(), because it contains an embedded unbalanced )

bugfix: lib/sh/local.sh: psql(): don't default the connection vars using use_local if running as the postgres user. in that case, connection must happen via a socket, with server="", and as the user running the command (postgres), with user="".

bugfix: lib/sh/db.sh: avoid outputting to /dev/fd/# when running as sudo on Linux, because this causes a "Permission denied" error (due to the /dev/fd/# file being owned by a different user). this is not a problem with normal redirects (>&#), because they do not use /dev/fd/# files which can have access permissions.

lib/sh/db.sh: pg_as_root(): run sudo with echo_run to help debug

bugfix: lib/sh/db.sh: pg_cmd(): only set PG* connection/login env vars when the corresponding var is non-empty. there are some situations in which these must be unset (in order to use the default value), and other situations when the var must be set to something (i.e. "") to avoid it being defaulted to a value in local.sh > connection vars.

bugfix: lib/sh/local.sh: pg_as_root(): need to use -E (preserve environment) option to sudo, so that $schema, $table get passed through

bugfix: lib/sh/local.sh: psql(): only \set schema, table if $schema, $table are non-empty, because otherwise, you will get a "zero-length delimited identifier" error

lib/sh/local.sh: added require_remote()

lib/sh/db.sh: added pg_as_root()

lib/sh/util.sh: added instructions for making an export only visible locally

bugfix: lib/sh/make.sh: don't allow rm to override remake if an invoked script uses this file. this fixes a bug in `rm=1 inputs/.../.../run` where the remake action would be invoked on the map_table command even though it had been suppressed, because it was run externally (i.e. make.sh was reloaded) and the rm=1 flag was still active

bugfix: lib/sh/util.sh: set_fds(): don't add surrounding quotes to empty redirect dest

bugfix: lib/sh/util.sh: set_fds(): need to check if redirect is empty before escaping it with `printf %q`, which may add surrounding quotes to an empty string

bugfix: lib/sh/util.sh: set_fds(): need to escape redirect destinations which are files, because they may contain special shell characters

lib/sh/util.sh: added rm_prefix()

lib/sh/db.sh: mysql_cmd(): added caller usage with connection/login opts

lib/sh/db.sh: mysql(), mysql_export(): usage: added database=...

lib/sh/db.sh: mk_select(): added support for ORDER BY

lib/sh/db.sh: added pg_export_table_to_dir(), analogous to pg_export_table_to_dir_no_header()

lib/sh/util.sh: is_array(): handle unset vars (=false). this fixes a bug in pg_export_table_no_header, which produced the error "lib/sh/util.sh: line 290: declare: cols: not found".

fix: lib/sh/util.sh: join(): documented that delim must be a single char

bugfix: lib/sh/db.sh: pg_dump(): don't default $struct flag to on, because both structure and data should be printed by default

lib/sh/db.sh: pg_dump(): added create_schema= flag to remove CREATE SCHEMA statements (useful if the schema already exists)

bugfix: lib/sh/util.sh: set_fds(): remove empty redirects resulting from using `redirs= cmd...` to clear the redirs and then using $redirs as an array

fix: lib/sh/util.sh: set_fds(): documented that it does not currently support redirecting an fd to itself (due to bash bugs that require the dest fd to be closed before it can be reopened)

bugfix: lib/sh/util.sh: stdout2fd(): don't add >&$fd redirect if the fd is 1, because redir does not currently support redirecting an fd to itself (due to bash bugs that require the dest fd to be closed before it can be reopened)

lib/sh/util.sh: filter_fd(): factored out >() subshell command into stdout2fd() for clarity

bugfix: lib/sh/util.sh: redir(): unset redirs so that you don't redirect again in the invoked command

fix: lib/sh/util.sh: filter_fd(): documented that ${redirs[@]} must not be set to an outer value

lib/sh/local.sh: added pg_dump_local()

lib/sh/db.sh: added pg_dump(), using the code in bin/pg_dump_vegbien with clarity improvements

lib/sh/db.sh: added pg_cmd() (analogous to mysql_cmd() for PostgreSQL), and use it in psql(), so that other PostgreSQL operations can use this to set the PG* connection/login vars

lib/sh/db.sh: pg_export(): added usage

lib/sh/db.sh: mysqldump(): added create_db=1 flag to print the CREATE DATABASE statement

bugfix: lib/sh/util.sh: pf(): echo func decl to stderr instead of stdout

lib/sh/util.sh: echo_vars(): documented that it only prints vars that are defined

lib/sh/util.sh: local_inv: also echo_vars the new var

lib/sh/util.sh: $sed_cmd: added usage

lib/sh/db.sh: mysqldump(): don't use --compatible=postgresql when the table structure is being exported, because this removes the table options (which include the COMMENT attribute). --compatible=postgresql remains on in data-only mode because embedded ` in data cannot easily be distinguished from ` around column names, so ANSI_QUOTES is needed to do the translation to " (and data sections do not contain table options). note that all --compatible modes that offer ANSI_QUOTES unfortunately exclude the table options, and there is no way to run a SQL query to set the SQL mode before beginning the dump, so ANSI_QUOTES translation must be handled by my2pg instead.

bugfix: lib/sh/util.sh: $sed_cmd: make output unbuffered, so that running e.g. bin/my2pg at the command line produces output as each line is read

lib/sh/db.sh: mk_select(): support passing $cols as array instead of SQL string, which is easier to enter in a shell script (less quotes, \ , etc.)

lib/sh/db.sh: added cols2list()

lib/sh/util.sh: added is_array()

lib/sh/db_make.sh: added pg_export_table_to_dir(), pg_export_tables_to_dir(). unlike db.sh pg_export_table_to_dir_no_header(), these functions are make-aware and will not clobber an existing file.

lib/sh/db.sh: psql(): display stack traces and DETAIL sections of error messages at verbosity 2+, to help debugging (previously they were always turned off). in particular, the DETAIL section of a "duplicate key value violates unique constraint" error is useful because it contains the duplicated key.

lib/sh/util.sh: moved runscript-related commands to lib/runscripts/util.run because these only apply to runscripts

bugfix: lib/sh/local.sh: prevent automated tests when the public schema contains the live DB, so the user doesn't have to explicitly specify can_test= when running the import on vegbiendev

bugfix: lib/sh/util.sh: sed: must use alias instead of function because function causes segfault in redir() subshell when used with make.sh make() filter (may be bug in bash?). this involves translating `unset LANG` to `env LANG=` (`env -u` to unset a var isn't supported on Mac, but fortunately sed treats LANG="" the same as unset LANG).

bugfix: lib/sh/util.sh: sed: unset LANG to avoid "illegal byte sequence" errors on invalid UTF-8 for LANG=*.UTF-8. these occur e.g. with MySQL data that is in Latin-1.

lib/sh/util.sh: sed: use function instead of alias so that env can be set up before calling sed

lib/sh/db.sh: psql(): hide the verbose CONTEXT information that is output with each NOTICE by setting the VERBOSITY psql var to terse (postgresql.1045698.n5.nabble.com/Quiet-quot-CONTEXT-quot-td1906036.html#a1906037)

*{.sh,run}: use new log-() instead of log+() with a negative #

lib/sh/util.sh: added log-() because it's non-obvious that you would otherwise have to invoke log+() with a negative #

bugfix: lib/sh/sync.sh: upload(): paths: don't dereference the path itself if it's a symlink; instead canonicalize just its parent dir. this allows syncing a specific file which is a symlink, rather than syncing the symlink's target.

lib/sh/util.sh: added canon_dir_rel_path(), which canonicalizes just the parent dir if the path is a symlink, to leave the symlink itself untouched

added lib/sh/datasrc.sh

lib/sh/make.sh: make(): run sys_cmd_path at a higher log_level since the make() steps should not be displayed by default

bin/make: moved $make_filter_active test to lib/sh/make.sh make() so that it's also used when make() is run directly (e.g. in a runscript) rather than via the bin/make wrapper in the PATH

bugfix: lib/sh/make.sh: make(): need to match absolute `make` paths such as /usr/bin/make

lib/sh/util.sh: added self_name alias and use it in self/self_sys

lib/sh/util.sh: added sys_cmd_path() and use it in cmd2sys

bugfix: lib/sh/util.sh: .rel(): first use realpath() on BASH_SOURCE¹ in case it's a symlink (as it is for bin/make)

bugfix: lib/sh/web.sh: curl(): use --fail so that curl returns a nonzero exit status on error (e.g. file not found) instead of appearing to exit successfully but outputting an error HTML document instead of the file

lib/sh/local.sh: added PATH_add= flag to allow turning off the addition of $bin_dir_abs to the PATH. this is useful for wrapper scripts that should not alter the PATH passed to their invoked command.

bugfix: lib/sh/make.sh: make(): invoke only the system make command instead of any wrapper for it in the PATH (by using self_sys instead of self), to prevent infinite recursion. single recursion is resolved by nonrecursive=1, but there are cases where mutual recursion occurs due to the presence of two, different bin/makes in the PATH (e.g. if you have two working copies with bin/make, and one is symlinked in your ~/bin/ folder), and these cases can only be resolved by clearing out the PATH completely (since the bin/makes do not know of each other's existence, in order to remove their parent dirs from the PATH).

lib/sh/util.sh: self_sys alias: use new sys_cmd() instead of `command -p` so that only the command path resolution is performed with a limited PATH, and the invoked command itself inherits the full PATH

lib/sh/util.sh: added sys_cmd(), which runs a system command and allows running a system command of the same name as the script

lib/sh/util.sh: added echo_builtin()

lib/sh/sync.sh: upload(): .rsync_filter: also support machine-specific filters, for cases when different machines produce the same file (e.g. a log file) but only one machine's copy should be backed up

lib/sh/local.sh: $sync_remote_url: allow user to override just the sync subdir (not the whole URL) in $sync_remote_subdir. this is useful e.g. for backing up the Mac's files to jupiter.

lib/sh/sync.sh: upload(): make put's $subpath option relative to the currdir instead, like the --include paths. note that $subpath unfortunately can't be used in subdirs at this point because it will cause rsync to ignore the .rsync_ignores and .rsync_filters in parent dirs, including the essential .rsync_ignore in the sync root dir.