selfDir_uZPPqC := $(dir $(lastword $(MAKEFILE_LIST)))

root := $(selfDir_uZPPqC)..

include $(root)/lib/common.Makefile

##### Configuration

# Command line

log ?= $(if $(test),,1)

profile ?=

quiet ?=

reverify ?= 1

use_staged ?= $(by_col)

# Makefile

exts ?= csv tsv txt xml

test_n ?= 2

##### Vars/functions

# Paths

datasrc := $(notdir $(realpath .))

bin := $(root)/bin

mappings := $(root)/mappings

# Make

SHELL := /bin/bash

selfMake = $(MAKE) --makefile=../input.Makefile

subMake = $(MAKE) $(@:$(root)/%=%) --directory=$(root)

+_ = $(+:_%=)

addBeforeExt = $(basename $(2))$(1)$(suffix $(2))

# OS

os := $(shell uname)

forOs = $(patsubst %,%-$(filter Linux Darwin,$(os)),$(1))

# Formatting

SED = sed -$(if $(filter Darwin,$(os)),E,r)

# System

date = $(shell date +"%Y-%m-%d-%H-%M-%S")

# Terminal

termCols := $(shell tput cols)

esc := '['

reset := $(esc)'0m'

emph := $(esc)'7m '

endEmph := ' '$(reset)

# Commands

MKDIR = mkdir -p

mkdir = $(MKDIR) $(@D)

CP = cp -p

diff = diff --unified=2

diffIgnoreSpace = $(diff) --ignore-space-change

diffVerbose = $(if $(verbose),diff --side-by-side --left-column\

--width=$(termCols),$(diff))

# BIEN commands

sortFilenames = $(shell $(bin)/sort_filenames $(1))

selfMap = $(bin)/cols 0 0

psqlOpts := --set ON_ERROR_STOP=1 --quiet

psqlAsBien := $(bin)/psql_vegbien $(psqlOpts)

# Usage: ($(inDatasrc); cat $(file))|$(psqlCmd)

inDatasrc = echo 'SET search_path TO "$(datasrc)";'

# SVN

addDir = $(if $(wildcard $(1)/),svn add --depth=empty $(1),svn mkdir $(1))

setSvnIgnore = svn propset svn:ignore $(2) $(1)

define addDirWithIgnore

$(addDir)

$(setSvnIgnore)

endef

##### Environment

export PATH := $(bin):$(PATH)

##### General targets

all: _always maps ;

clean: _always

	$(RM) $(all)

remake: _always clean

	+$(selfMake)

# re-run make so that cache of existing files is reset

make_script = ./$< >$@

%/: % _always ;

%: %.make

	$(if $(wildcard $@),,"time" $(make_script))

# Only remake if doesn't exist. This prevents unintentional remaking when the

# make script is newly checked out from svn (which sets the mod time to now) but

# the output is synced externally.

# Can't remove prereq to do this, because it determines when the rule applies.

.PRECIOUS: % # save partial outputs of aborted src make scripts

##### SVN

add: _always

	$(call setSvnIgnore,.,'*')

	$(call addDirWithIgnore,logs,$$'*.log.sql\n*.trace')

	$(call addDirWithIgnore,verify,'*.out')

	$(call addFile,import_order.txt)

# Adds a new table subdir

%/add: _always

	$(call addDirWithIgnore,$*,'*')

	$(call addDirWithIgnore,$*/logs,$$'*.log.sql\n*.trace')

##### Existing maps discovery

sortFile := import_order.txt

tables := $(if $(wildcard $(sortFile)),$(shell cat $(sortFile)))

    # $(shell) replaces "\n" with " "

allSubdirs := $(call wildcard/,*/)

allTables := $(call sortFilenames,$(filter-out _% verify logs,$(allSubdirs:%/=%)))

joinedTables := $(filter-out $(tables),$(allTables))

allTables := $(joinedTables) $(tables)# move joined tables to beginning

ifeq ($(tables),)# none specified in sort file

tables := $(allTables)

endif

srcMaps := $(wildcard */src.csv)

srcMap := %/src.csv

viaMap := %/map.csv

fullViaMap := %/map.full.csv

directMap := %/VegBIEN.csv

anyMap := $(srcMap) $(viaMap) $(fullViaMap) $(directMap) %/unmapped_terms.csv\

%/new_terms.csv

extsFilter := $(addprefix %.,$(exts))

dataOnly = $(filter $(extsFilter),$(1))

anyTest = $*/test.%

srcsOnly = $(filter-out $(anyMap) $(anyTest) %/logs,$(call dataOnly,$(1)))

via := Veg+

coreMap := $(mappings)/$(via)-VegBIEN.csv

coreSelfMap := $(mappings)/$(via).self.csv

vocab := $(mappings)/$(via).vocab.csv

dict := $(mappings)/$(via)-VegCore.csv

viaMaps := $(wildcard $(tables:%=%/map.csv))

viaMaps += $(filter-out $(viaMaps),$(srcMaps:%/src.csv=%/map.csv))

viaMaps += $(filter-out $(viaMaps) $(fullViaMap),$(wildcard */map.csv))

autogenMaps := $(subst map.,VegBIEN.,$(viaMaps))

directMaps := $(autogenMaps) $(filter-out $(autogenMaps),\

$(wildcard */VegBIEN.csv))

##### Sources

srcs = $(call sortFilenames,$(call srcsOnly,$(wildcard $*/*)))

nonHeaderSrcs = $(filter-out %/header.csv,$(srcs))

isRef = $(if $(nonHeaderSrcs),,1)

    # empty subdir, so references an already-installed staging table

isXml = $(filter %.xml,$(nonHeaderSrcs))

nonXml = $(if $(isXml),,1)

isCsv = $(if $(nonHeaderSrcs),$(if $(isXml),,1))

    # true if $(srcs) non-empty and contains no *.xml

catSrcs = $(bin)/cat$(if $(nonXml),_csv) $(srcs)

withCatSrcs = $(catSrcs:$(bin)/%=$(bin)/with_%) --

# Usage: `make {--silent|-s} inputs/<datasrc>/cat` (don't echo make commands)

cat: $(tables:%=%/cat) _always ;

%/cat: _always

	$(catSrcs)

##### Staging tables installation

dbExports := $(sort $(wildcard *.schema.sql))# schemas first

dbExports += $(sort $(filter-out $(dbExports),$(wildcard *.sql)))# all others

dbExports := $(strip $(dbExports))# += adds extra whitespace

allInstalls := $(if $(dbExports),sql) $(allTables)

install: _always schema $(allInstalls:%=%/install) ;

uninstall: _always rm_schema ;

# rm_schema will also drop all staging tables

reinstall: _always uninstall install ;

schema: _always

	-echo 'CREATE SCHEMA "$(datasrc)";'|$(psqlAsBien)

# ignore errors if schema exists

rm_schema: _always

	echo 'DROP SCHEMA IF EXISTS "$(datasrc)" CASCADE;'|$(psqlAsBien)

logInstall = $(if $(log),$(if $(quiet),&>,2>&1|tee $(2)\

)$(1)logs/install.log.sql)

logInstall* = $(call logInstall,$*/)

logInstall*Add = $(call logInstall,$*/,-a)# append to log

# Must come before %/install to override it

sql/install: $(dbExports)

	($(inDatasrc); cat $+|grep -vF 'SET search_path')|"time" $(psqlAsBien) \

$(logInstall)

# The export must be created with:

# `--compatible=postgresql --add-locks=false --set-charset --no-create-info`

# Must come before `%.sql: _MySQL/%.sql` to override it

%.data.sql: _MySQL/%.data.sql

	$(bin)/my2pg.data <$< >$@

# The export must be created with:

# `--compatible=postgresql --add-locks=false --set-charset`

# Add `--no-data` to create a schema-only export.

%.sql: _MySQL/%.sql

	$(bin)/my2pg <$< >$@

cleanup = (prefix=; . $(bin)/vegbien_dest; env schema=$(datasrc) table=$*\

$(bin)/csv2db) $(logInstall*Add)

define exportHeader

$(cleanup)

echo 'SELECT * FROM "$(datasrc)"."$*" LIMIT 0;'|$(psqlAsBien) \

--no-align --field-separator=, --pset=footer=off >$*/header.csv

endef

# For staging tables which are derived by joining together other staging tables.

%/install %/header.csv: %/create.sql _always

	($(inDatasrc); echo 'CREATE TABLE "$*" AS'; cat $<; echo ';')|"time" \

$(psqlAsBien) --echo-all --set=table='"$*"' $(logInstall*)

	-$(addRowNum)

	$(exportHeader)

# ignore errors if create.sql already added a primary key

addRowNum = echo 'ALTER TABLE "$(datasrc)"."$*"\

ADD COLUMN row_num serial NOT NULL PRIMARY KEY;'|$(psqlAsBien) $(logInstall*Add)

# The joined tables must be suffixed with ".src" to prevent the creation of a

# row_num column, which collides during joins.

isJoinedTable = $(filter %.src,$*)

hasRowNum = $(if $(isJoinedTable),,1)

%/install: _always

	$(if $(isRef),$(exportHeader),$(if $(nonXml),$(import_install_)))

import_install_ = (prefix=; . $(bin)/vegbien_dest; "time" nice -n +5\

env schema=$(datasrc) table=$* has_row_num=$(hasRowNum) $(bin)/csv2db\

$(catSrcs) $(logInstall*))

##### Maps building

# WARNING: You CANNOT make a subdir using `make inputs/<datasrc>/<subdir>/`.

# You must instead make the entire datasource dir: `make inputs/<datasrc>/`

# Maps to (try to) build are added to one of these

maps :=

createOnlyMaps :=

srcRoot = $(mappings)/root.sh

mkSrcMap = $(catSrcs)|(. $(srcRoot); env datasrc=$(datasrc) $(bin)/src_map >$@)

# Src maps cleanup

ifneq ($(filter %/.src.csv.last_cleanup,$(MAKECMDGOALS)),)

%/.src.csv.last_cleanup: %/src.csv

	$(bin)/in_place $< $(bin)/cols +

	touch $@

else

# Autogen src maps with known table names

%/src.csv: _always

	$(if $(wildcard $@),,$(if $(nonXml),$(mkSrcMap)))

	+$(if $(nonXml),$(selfMake) $(@:%/src.csv=%/.src.csv.last_cleanup))

# only build if doesn't exist

# only build if CSV srcs exist for that table name

endif

createOnlyMaps += $(tables:%=%/src.csv)

# Must come before $(root)/% to override it

$(coreSelfMap): _always

	-+$(subMake)

# ignore errors if $(coreSelfMap) does not exist

# Via maps cleanup

ifneq ($(filter %/.map.csv.last_cleanup,$(MAKECMDGOALS)),)

%/.map.csv.last_cleanup: %/map.csv $(coreMap) $(vocab) $(dict)

	$(bin)/in_place $< $(bin)/canon 1 $(word 3,$+)

	$(bin)/in_place $< $(bin)/translate 1 $(word 4,$+)

	$(bin)/in_place $< env ignore=1 $(bin)/subtract $(word 2,$+) 0 1 2

	touch $@

	+$(selfMake) $(<:%/map.csv=%/unmapped_terms.csv)

	+$(selfMake) $(<:%/map.csv=%/new_terms.csv)

# Include comment column so commented mappings are never removed

else

%/map.csv: %/src.csv _always

	$(if $(wildcard $@),,$(if $(wildcard $<),$(CP) $< $@))

	+$(selfMake) $(@:%/map.csv=%/.map.csv.last_cleanup)

endif

srcMap* = $*/src.csv

joinSrcMap = $(if $(hasSrc),$(bin)/in_place $@ $(bin)/intersect $(srcMap*) 0)

hasSrc = $(shell test -s $(word 3,$+) && echo t)

makeFullCsv = env ignore=1 $(bin)/union <$(wordlist 1,2,$+)|$(bin)/sort_map >$@

%/map.full.csv: %/map.csv $(coreSelfMap)

	$(makeFullCsv)

	$(joinSrcMap)

maps += $(patsubst %/map.csv,%/map.full.csv,$(viaMaps))

%/VegBIEN.csv: %/map.full.csv $(coreMap)

	$(bin)/join <$+|$(bin)/sort_map >$@

maps += $(autogenMaps)

maps: $(createOnlyMaps) $(maps) _always ;

all += $(maps)

##### Maps validation

define newTerms

$(bin)/filter_out_ci <$< $(1) $(word 2,$+)|$(bin)/cols $(1)$(2)|tail -n +2 >$@

$(bin)/autoremove $@

endef

%/unmapped_terms.csv: %/map.csv $(vocab)

	$(call newTerms,1)

%/new_terms.csv: %/map.csv $(vocab) %/unmapped_terms.csv

	$(call newTerms,0,$(rmNewTerms))

rmNewTerms = $(if $(wildcard $(word 3,$+)),|$(bin)/filter_out_ci 0 $(word 3,$+))

termsSubdirs := $(tables)

include $(root)/lib/mappings.Makefile

##### External dependencies

$(root)/%: _always

	+$(subMake)

.PRECIOUS: $(root)/% # let ext. dir's Makefile decide whether to delete on error

##### Mapping

+maps = $(filter %/map.csv %/VegBIEN.csv $(mappings)/%,$(+_))

map = $(if $(srcs),$(withCatSrcs) $(root)/map $(+maps),\

$(shell echo Warning: No input file $*/*.{$(exts)} >&2)false)

# need false to run some command, when prefixed by `env ...` below

map2db = $(if $(use_staged),env in_database=vegbien in_schema=$(datasrc)\

in_table=$*) env out_database=vegbien $(map)

##### Import to VegBIEN

profileTest = $(if $(profile),$(if $(test),1))

profileOnly = -env profile_to=/dev/fd/3 $(map2db) 3>&1 1>&2|\

$(bin)/profile_stats /dev/fd/0

log_ = $*/logs/$(if $(n),n=$(n).,)$(date).log.sql

trace = $(log_:.log.sql=.trace)

import = -$(if $(profileTest),$(profileOnly),(set -x; "time" env commit=1\

$(if $(profile),profile_to=$(trace)) $(map2db)) $(if $(log),\

$(if $(n),,&>$(log_)))$(if $(log),$(if $(n), 2>&1|tee -a $(log_))))

# don't abort on import errors, which often relate to invalid input data

import: $(tables:%=%/import) _always ;

%/import: %/VegBIEN.csv _always

	$(import)

# default:

%/import: _always ;

##### Log files from import

logs := $(wildcard */logs/*.log.sql */logs/*.trace)

rm_logs: _always

	$(RM) $(logs)

##### Verification of import

verify: $(tables:%=%/verify) _always ;

%/verify: verify/%.ref verify/%.out _always

	-$(diffVerbose) $(+_)

# don't abort on verification errors, which are expected during development

# default:

%/verify: verify/%.out _always

	$(if $(shell test -e $< && echo t),cat $<)

# don't run if verify/%.out's default do-nothing action was used

# can't use $(wildcard) because it won't recheck file after verify/%.out is run

psqlExport := "time" $(psqlAsBien) --no-align --field-separator=$$'\t'\

--pset=footer=off --pset=null=NULL

verify = $(if $(reverify),$(psqlExport) --set=datasource="'$(datasrc)'" <$< >$@)

verify/%.out: $(mappings)/verify.%.sql _always

	$(verify)

# default:

verify/%.out: _always ;

all += $(wildcard verify/*.out)

%.ref: %.ref.sql

	($(inDatasrc); cat $<)|$(psqlExport) >$@

##### Editing import

rotate: _always

	echo "UPDATE party SET organizationname = organizationname||'.$(date)'\

WHERE organizationname = '$(datasrc)';"|$(psqlAsBien)

rm: _always

	echo "DELETE FROM party WHERE organizationname = '$(datasrc)';"|\

$(psqlAsBien)

##### Testing

testRefOutput = $(subst .by_col,,$(1))

testRef = $(testRefOutput).ref

hasOwnRef = $(filter $@,$(call testRefOutput,$@))

# filter returns non-empty if they are equal

# `rm $@`: Remove outputs of successful tests to reduce clutter

# `$(foreach use_staged...)`: Run with use_staged=1

define runTest

@echo "Testing $(abspath $@)..."

>$@ env test=1 n=$(test_n) $(1) $(foreach use_staged,1,$(map2db))

@(set -x; $(diffIgnoreSpace) $(call testRef,$@) $@) 2>&1 && rm $@ || { e=$$?;\

$(if $(wildcard $(call testRef,$@)),,cat $@;)\

$(if $(hasOwnRef),\

{\

read -p $(emph)'Accept new test output? (y/n)'$(endEmph) REPLY;\

if test "$$REPLY" = y; then\

(set -x; $(MAKE) $@-ok --directory=$(realpath .) --makefile=../input.Makefile);\

exit 0;\

fi;\

};,\

echo $(emph)"Note: The preceding failed test is compared to another test's\

output"$(endEmph);\

echo $(emph)"When it fails, this always indicates a bug"$(endEmph);\

)\

exit $$e;}

endef

tests :=

# Requires staging tables. To create them, run `make inputs/<datasrc>/install`.

# Non-flat-file inputs fall back to mimicking a successful test

%/test.xml: %/VegBIEN.csv _always

	$(if $(nonXml),$(call runTest,by_col=))

tests += %/test.xml

%/test.by_col.xml: %/VegBIEN.csv _always

	-$(if $(nonXml),$(call runTest,by_col=1))

# Don't abort tester if only by-column test fails

# Only run column-based tests if column-based mode enabled, because these tests

# are much slower than the row-based tests for small numbers of rows

ifneq ($(by_col),)

tests += %/test.by_col.xml

endif

testOutputs := $(foreach test,$(tests),$(tables:%=$(test)))

.PRECIOUS: $(testOutputs) # save outputs of failed tests so they can be accepted

test: _always $(testOutputs) ;

all += $(wildcard %/test*.xml)

# Accepts a test output: make <test_output_path>-ok

%-ok: _always

	mv $* $(call testRef,$*)

accept-all: _always

	+yes|$(selfMake) test

##### Documentation

steps = $(selfMake) -s $*/import test=1 by_col=1 verbosity=2 n=100\

2>&1|$(bin)/debug2redmine >$@

%/logs/steps.by_col.log.sql: _always

	+$(steps)