/ Docs

Sections — Company Map

In DartLab, sections is the entire map of a company. It takes disclosure documents filed each reporting period, splits them by topic, and aligns them side by side across time into a single horizontalized board. Instead of opening individual annual reports one at a time, you can compare every disclosure section chronologically at a glance.

What is sections

Korean DART annual reports contain dozens of sections — from Chapter I (Company Overview) to Chapter XII (Detailed Schedules). The problem is that each report exists as a separate file per quarter/year. The same topic (“Business Overview”) is scattered across 2025, 2024, 2024Q3, and so on.

sections restructures these scattered originals into a topic × period matrix. Each row is one topic-block, and each column is a period. Text and tables are preserved as-is from the original, just aligned on the time axis.

import dartlab

c = dartlab.Company("005930")  # Samsung Electronics
c.sections

The returned DataFrame looks like this:

chapter │ topic            │ blockType │ textNodeType │ 2025Q4 │ 2024Q4 │ 2024Q3 │ …
I       │ companyOverview  │ text      │ heading      │ "…"    │ "…"    │ "…"    │
I       │ companyOverview  │ text      │ body         │ "…"    │ "…"    │ "…"    │
I       │ companyOverview  │ table     │ null         │ "…"    │ "…"    │ null   │
II      │ businessOverview │ text      │ heading      │ "…"    │ "…"    │ "…"    │

Key Columns

Structure Columns

ColumnDescription
chapterTop-level chapter number. I through XII matching the annual report structure
topicStandardized topic identifier. snakeCase names like companyOverview, businessOverview, BS, IS, CF
blockType"text" or "table". Distinguishes narrative blocks from table blocks within the same topic
blockOrderBlock order within a topic. Preserves the original document sequence
textNodeTypeSub-type for text blocks: "heading" (section title) or "body" (narrative text). null for tables
textLevelHeading depth level (1, 2, 3, …). null for body and table blocks
textPathStructural path of the heading. Indicates position within the section

Period Columns

Columns like 2025Q4, 2024Q4, 2024Q3, 2024Q2, 2024Q1, … follow in order. Newest period comes first. Annual reports are labeled as Q4.

  • Each cell contains the original payload for that period
  • Text blocks contain narrative text; table blocks contain markdown tables
  • null means no data exists for that period

show(topic) — Open a Topic

After exploring sections, use show() when you want to dive deeper into a specific topic.

c.show("overview")       # block index DataFrame (block, type, source, preview)
c.show("companyOverview", 0)    # actual data for block 0
c.show("BS")                    # balance sheet (finance source)

show() always operates on top of sections, with source priority:

  1. finance (BS, IS, CF, CIS, SCE) — numbers are authoritative, so they override docs text
  2. report — DART structured disclosure data
  3. docs — original narrative text/tables

Finance topics return normalized numeric DataFrames from the finance engine. Narrative topics return the original text aligned by period.

trace(topic) — Check Source

To see which source was actually selected, use trace().

c.trace("BS")               # finance source, per-period coverage
c.trace("overview")  # docs source

The result includes:

  • Selected source (docs / finance / report)
  • Per-period data availability (coverage)
  • Metadata like chapter, label

Filtering sections

sections is a Polars DataFrame, so you can filter freely with Polars syntax.

import polars as pl

# specific topic only
df = c.sections.filter(pl.col("topic") == "companyOverview")

# text blocks only
df = c.sections.filter(pl.col("blockType") == "text")

# table blocks only
df = c.sections.filter(pl.col("blockType") == "table")

sections also provides convenience methods:

c.sections.periods()    # list of periods
c.sections.ordered()    # sorted newest first
c.sections.coverage()   # per-topic period coverage summary

docs.sections vs sections

Currently c.sections and c.docs.sections are identical. Both point to the pure docs source view.

c.sections              # shortcut
c.docs.sections         # explicit path (same result)

show(topic) applies source priority (finance > report > docs) to return a merged result, so for practical use, just use show().

Same for EDGAR

US SEC EDGAR companies use the same structure and same API.

us = dartlab.Company("AAPL")
us.sections
us.show("10-K::item1Business")
us.show("BS")

Only the topic names differ, following SEC form conventions. 10-K Item 1 is 10-K::item1Business, Item 7 is 10-K::item7MDA, and so on. Financial statement topics (BS, IS, CF) use the same names as DART.

Next Steps