Files
simple-aircraft-manager-rails/app/services/oil_report_parser/avlab.rb
T
rickyandClaude Opus 5.5 1c6c2f88c0 Blackstone oil analysis, .sam.zip export/import, AI logbook transcription
- Oil analysis: Blackstone PDF import (20 elements, oil properties, universal
  averages), results stored per element, oil type, "leave out of trends",
  the original app's statuses; AvLab fixes (multi-line comments, unrelated
  labels no longer steal values). Existing records migrate.
- Aircraft archives in the original app's .sam.zip format (schema v2): export
  from Access & Sharing, background import from the Fleet page, rake tasks.
  Archives move both ways between this app and the original.
- Scanned logbook pages (images or PDFs) upload as a document; with
  ANTHROPIC_API_KEY set, Claude transcribes them into logbook entries.
- Fix: deleting a component or aircraft failed when AD compliances or
  inspection records pointed at a component.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-11-06 21:05:19 -05:00

209 lines
8.4 KiB
Ruby
Raw Blame History

This file contains ambiguous Unicode characters
This file contains Unicode characters that might be confused with other characters. If you think that this is intentional, you can safely ignore this warning. Use the Escape button to reveal them.
class OilReportParser
# Aviation Laboratories (AvLab) report. Each sample block has its dates,
# hours, the lab's verdict, comments and a ppm table (value, lab average beneath).
#
# The PDF's text layer is messy: it hides the cell color codes ("#00CC00") as
# larger text on top of the results. So this works from positioned characters,
# drops the color-code layer by font size, rebuilds words and lines, and maps
# each result to its element column by x position.
class Avlab
LAB = "AvLab".freeze
COLOR_CODE_FONT_SIZE = 8.5 # the hidden color codes are 9pt; real text is 7–8pt
WORD_GAP = 1.0 # glyphs closer than this are one word (spaces run 1.3pt+)
COLUMNS = {
"Iron" => "iron", "Copper" => "copper", "Nickel" => "nickel", "Chromium" => "chromium",
"Silver" => "silver", "Magnesium" => "magnesium", "Aluminum" => "aluminum", "Lead" => "lead",
"Silicon" => "silicon", "Titanium" => "titanium", "Tin" => "tin", "Moly." => "molybdenum"
}.freeze
LABELS = {
"Sample Date:" => :sample_date, "Analysis Date:" => :analysis_date, "Sample Number:" => :sample_number,
"TSN/TSO:" => :engine_hours, "Oil Hours:" => :oil_hours, "Filter Hours:" => :filter_hours,
"Oil Added:" => :oil_added
}.freeze
LABEL_MAX_X = 108 # left-hand labels end before here…
VALUE_X = 105..160 # …and their values sit in this column
Char = Struct.new(:text, :x, :endx, :y, :page)
Word = Struct.new(:text, :x, :endx, :y, :page) do
def center = (x + endx) / 2.0
end
Line = Struct.new(:y, :page, :words) do
def text = words.map(&:text).join(" ")
end
def self.from_receivers(receivers)
chars = receivers.each_with_index.flat_map do |receiver, index|
# Keep overlapping glyphs: a real "0" under a hidden one would otherwise be dropped.
receiver.runs(merge: false, skip_overlapping: false).filter_map do |run|
next if run.font_size >= COLOR_CODE_FONT_SIZE || run.text.strip.empty?
Char.new(run.text, run.x, run.endx, run.y, index)
end
end
new(chars).parse
end
def initialize(chars)
@lines = build_lines(chars)
end
def parse
unless @lines.any? { _1.text.include?("OIL ANALYSIS RESULTS IN PARTS PER MILLION") }
raise Error, "Not an AvLab oil analysis report"
end
samples = sample_blocks.map { parse_sample(_1) }.select { _1[:sample_date] }
raise Error, "No samples found in the report" if samples.empty?
Report.new(lab: LAB, tail_number: after("Tail No.:"), engine_serial: after("Engine S/N:"),
engine_model: after("Engine Model:", words: 3), samples: samples)
end
private
# Characters → lines (same baseline ±1pt) → words (glyphs closer than a space).
def build_lines(chars)
lines = []
chars.sort_by { [ _1.page, -_1.y ] }.each do |c|
line = lines.reverse_each.find { _1.page == c.page && (_1.y - c.y).abs <= 1.0 }
line ? line.words << c : lines << Line.new(c.y, c.page, [ c ])
end
lines.each do |line|
line.words = line.words.sort_by(&:x).each_with_object([]) do |c, words|
last = words.last
if last && c.x - last.endx <= WORD_GAP
last.text += c.text
last.endx = c.endx
else
words << Word.new(c.text.dup, c.x, c.endx, c.y, c.page)
end
end
end
end
def header_lines
first = @lines.find { _1.words.first&.text.in?(%w[CURRENT PREVIOUS]) }
first ? @lines.take_while { _1 != first } : @lines
end
# The word(s) following a label on the same line, e.g. "Tail No.:" → "N12345".
def after(label, words: 1)
parts = label.split
header_lines.each do |line|
texts = line.words.map(&:text)
texts.each_cons(parts.size).with_index do |window, i|
next unless window == parts
found = texts[i + parts.size, words]&.take_while { !_1.end_with?(":") }
return found.join(" ").presence if found&.any?
end
end
nil
end
# Lines from each "CURRENT SAMPLE" / "PREVIOUS SAMPLE n" heading to the next.
def sample_blocks
starts = @lines.each_index.select { |i| @lines[i].words.first&.text.in?(%w[CURRENT PREVIOUS]) && @lines[i].words[1]&.text == "SAMPLE" }
starts.each_with_index.map { |s, n| @lines[s...(starts[n + 1] || @lines.size)] }
end
def parse_sample(block)
heading = block.first
verdict = heading.words.select { _1.x > 140 && _1.x < 470 }.map(&:text).join(" ")
fields = labeled_values(block)
comments = comments(block)
elements, averages = results(block)
{
current: heading.words.first.text == "CURRENT",
sample_date: OilReportParser.date(fields[:sample_date]),
analysis_date: OilReportParser.date(fields[:analysis_date]),
sample_number: fields[:sample_number],
engine_hours: OilReportParser.number(fields[:engine_hours]),
oil_hours: OilReportParser.number(fields[:oil_hours]),
filter_hours: OilReportParser.number(fields[:filter_hours]),
oil_added: OilReportParser.number(fields[:oil_added]),
status: status(verdict),
lab_comments: [ (verdict unless status(verdict) == "normal"), comments ].compact_blank.join("\n").presence,
elements: elements,
element_averages: averages
}
end
# Left-hand "Label:" lines and the values beside them. A value isn't always on
# the label's baseline, so each value goes to the nearest label above or below.
# Every label counts, including ones we don't keep (e.g. "Filter Wt. (mgs):"),
# so their values don't drift onto a neighbouring label.
def labeled_values(block)
labels = block.filter_map do |line|
text = line.words.select { _1.x < LABEL_MAX_X - 20 || (_1.x < LABEL_MAX_X && _1.text.end_with?(":")) }.map(&:text).join(" ")
[ LABELS[text], line.y ] if text.end_with?(":")
end
values = block.drop(1).flat_map { |line| line.words.select { VALUE_X.cover?(_1.x) } } # skip the heading
values.each_with_object({}) do |word, found|
key, = labels.min_by { (_2 - word.y).abs }
found[key] ||= word.text if key
end
end
# "Comments: …" and the lines that continue it at the same left margin.
def comments(block)
start = block.index { _1.words.first&.text == "Comments:" } or return nil
first = block[start]
margin = first.words.first.x
lines = [ first.words.drop(1) ]
block[(start + 1)..].each_with_index do |line, i|
previous = block[start + i]
break unless (line.words.first.x - margin).abs <= 3 && previous.y - line.y <= 14 && line.page == previous.page
lines << line.words
end
lines.flatten.map(&:text).join(" ").presence
end
# The ppm row under the element headings, and the (average) row under that.
def results(block)
elements, averages = {}, {}
header = block.find { |line| COLUMNS.keys.count { line.words.map(&:text).include?(_1) } >= 6 } or return [ elements, averages ]
columns = header.words.filter_map { |w| [ COLUMNS[w.text], w.center ] if COLUMNS[w.text] }
below = block.select { _1.y < header.y && _1.y > header.y - 25 }
below.each do |line|
# Results start right of the label column ("Cylinder Type: steel" shares these lines).
merge_less_than(line.words.select { _1.x >= VALUE_X.end }).each do |word|
element, = columns.min_by { (_2 - word.center).abs }
if word.text.start_with?("(")
averages[element] = ppm(word.text.delete("()"))
else
elements[element] = ppm(word.text)
end
end
end
[ elements.compact, averages.compact ]
end
# "<" and "0.1" are drawn as separate words.
def merge_less_than(words)
words.each_with_object([]) do |w, out|
if out.last&.text == "<"
out.last.text = "< #{w.text}"
out.last.endx = w.endx
else
out << w.dup
end
end
end
# "N/A" → nil; "< 0.1" (below detection) → 0; ".3" → 0.3
def ppm(text)
return nil if text.blank? || text == "N/A"
return 0 if text.start_with?("<")
OilReportParser.number(text)
end
def status(verdict)
case verdict.upcase
when /ABNORMAL|HIGH|CRITICAL|SEVERE|PROBLEM/ then "action_required"
when /APPEARS NORMAL/ then "normal"
else "monitor"
end
end
end
end