# One year on, 2024 to 2025 **Answer.** Between the 2024 and 2025 declarations Uruguay gained **127,121 cattle** (+1.05%) and lost **621,595 sheep** (-11.47%), while the number of holders declaring beef cattle went from 61,561 to 59,121 (-4.0%): more cattle, fewer hands. Notebook: https://cow-tools.sophia-labs.com/notebooks/year-over-year/year-over-year.ipynb · Page: https://cow-tools.sophia-labs.com/notebooks/year-over-year/index.html · Data: Uruguay MGAP/SNIG DICOSE 2024–2025 (graph `uruguay-cows` at https://cow-tools.sophia-labs.com/mcp). ## Findings - Between the 2024 and 2025 declarations Uruguay gained **127,121 cattle** (+1.05%) and lost **621,595 sheep** (-11.47%), while the number of holders declaring beef cattle went from 61,561 to 59,121 (-4.0%): more cattle, fewer hands. - Growth came from the north and centre (**Salto** +5.9%, Durazno +5.8%, Rivera +5.2%); the south and east shrank (**Canelones** -4.8%, Treinta y Tres -4.6%). ## The notebook Two declarations, a year apart. Before comparing them we read what the cube says about its own quality: cow-tools records every warning raised at ingest as a `cow:QualityWarning` in the year's quality graph. A comparison that leans on a dimension a warning touches must say so, or not be made. ```python # Setup. import re import altair as alt import pandas as pd import cowtools_client as cw import cowtools_charts as cc names = cc.department_names() ``` ## First, the warnings ```python # The query: every quality report and warning the cube recorded, per year. reports = cw.sparql(""" SELECT ?year ?passed ?blocking ?warnings WHERE { GRAPH ?g { ?r a cow:QualityReport ; cow:sourceYear ?year ; cow:passed ?passed ; cow:blockingViolationCount ?blocking ; cow:warningCount ?warnings } } ORDER BY ?year """) warnings = cw.sparql(""" SELECT ?year ?message WHERE { GRAPH ?g { ?w a cow:QualityWarning ; cow:sourceYear ?year ; cow:message ?message } } ORDER BY ?year ?message """, allow_empty=True) cc.show_table(reports, name="quality-reports") cc.show_table(warnings, name="quality-warnings") ``` | year | passed | blocking | warnings | |:-------|:---------|-----------:|-----------:| | 2,024 | True | 0 | 1 | | 2,025 | True | 0 | 3 | | year | message | |:-------|:----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------| | 2,024 | optional source file not found: Codigos de areas de enumeracion | | 2,025 | codelist 'scheme_mgap_production_types' has no table for snapshot year 2025; borrowed 9 concept(s) from snapshot year 2024 (explicit cross-year borrow, recorded as ) | | 2,025 | optional source file not found: Codigos de areas de enumeracion | | 2,025 | optional source file not found: Codigos de tipos de produccion MGAP | ```python cw.check("both years passed their quality gate", bool(reports["passed"].all()), True, source="cow:QualityReport") cw.check("no blocking violations in either year", int(reports["blocking"].sum()), 0, source="cow:QualityReport") cw.check("every counted warning is recorded", int(reports["warnings"].sum()), len(warnings), source="cow:warningCount vs cow:QualityWarning rows") ``` ```text PASS both years passed their quality gate: True vs True (exactly) — cow:QualityReport PASS no blocking violations in either year: 0 vs 0 (exactly) — cow:QualityReport PASS every counted warning is recorded: 4 vs 4 (exactly) — cow:warningCount vs cow:QualityWarning rows ``` Each warning names what it touches: a codelist borrowed from another year, or an optional source file missing from a package. This comparison uses three things: the **department** and **species** codelists and the **datosanimales** table. We collect what the warnings touch and refuse to go on if any of the three is among them. ```python touched = set() for msg in warnings["message"]: touched |= set(re.findall(r"codelist '([^']+)'", msg)) touched |= set(m.strip() for m in re.findall(r"optional source file not found: (.+)", msg)) print("touched by a warning:", sorted(touched) or "nothing") uses = {"departments", "species", "datos_animales", "Datos animales"} hit = sorted(t for t in touched if any(u.lower().replace("_", " ") in t.lower().replace("_", " ") for u in uses)) cw.check("no warning touches the dimensions this comparison uses", hit, [], source=f"uses: {sorted(uses)}") ``` ```text touched by a warning: ['Codigos de areas de enumeracion', 'Codigos de tipos de produccion MGAP', 'scheme_mgap_production_types'] PASS no warning touches the dimensions this comparison uses: [] vs [] (exactly) — uses: ['Datos animales', 'datos_animales', 'departments', 'species'] ``` ## Cattle, sheep and holders, both years ```python # The query: stock and holders per department, species and year. q = cw.sparql(""" SELECT ?year ?department ?species (SUM(?n) AS ?animals) (SUM(?h) AS ?holders) WHERE { GRAPH ?g { ?obs a cow:AnimalFlowObservation ; cow:year ?year ; cow:department ?department ; cow:species ?species ; cow:animalsCount ?n ; cow:holdersCount ?h . } } GROUP BY ?year ?department ?species """) q["code"] = q["department"].map(cw.dept_code) q["sp"] = q["species"].map(cw.dept_code) q["kind"] = q["sp"].map({"1": "cattle", "11": "cattle", "2": "sheep"}) nat = q.dropna(subset=["kind"]).groupby(["kind", "year"])["animals"].sum().unstack() beef_holders = q.query("sp == '1'").groupby("year")["holders"].sum() nat ``` ```text year 2024 2025 kind cattle 12138148 12265269 sheep 5420661 4799066 ``` ```python for year in (2024, 2025): csv = cw.source_csv(year, "datosanimales.csv") beef = csv[csv["EspecieCodigo"] == "1"] cw.check(f"{year} holders declaring beef cattle: cube = CSV", int(beef_holders[year]), int(cw.num(beef["CantidadTenedoresPorEspecie"]).sum()), source=f"datosanimales.csv {year}, CantidadTenedoresPorEspecie") change = nat[2025] / nat[2024] - 1 cw.check("cattle 2024 → 2025", round(change["cattle"], 4), 0.0105, tol=0.00005, source="expedition 2026-07-05: +1.05%") cw.finding("headline", f"Between the 2024 and 2025 declarations Uruguay gained **{int(nat.loc['cattle', 2025] - nat.loc['cattle', 2024]):,} cattle** " f"({change['cattle']:+.2%}) and lost **{int(nat.loc['sheep', 2024] - nat.loc['sheep', 2025]):,} sheep** " f"({change['sheep']:+.2%}), while the number of holders declaring beef cattle went from " f"{int(beef_holders[2024]):,} to {int(beef_holders[2025]):,} ({beef_holders[2025]/beef_holders[2024]-1:+.1%}): " f"more cattle, fewer hands.") cw.check("fewer holders declared beef cattle in 2025 (so 'fewer hands' holds)", bool(beef_holders[2025] < beef_holders[2024]), True, source="the two numbers above") ``` ```text PASS 2024 holders declaring beef cattle: cube = CSV: 61,561 vs 61,561 (exactly) — datosanimales.csv 2024, CantidadTenedoresPorEspecie PASS 2025 holders declaring beef cattle: cube = CSV: 59,121 vs 59,121 (exactly) — datosanimales.csv 2025, CantidadTenedoresPorEspecie PASS cattle 2024 → 2025: 0.0105 vs 0.0105 (within ±5e-05) — expedition 2026-07-05: +1.05% ``` > **Finding.** Between the 2024 and 2025 declarations Uruguay gained **127,121 cattle** (+1.05%) and lost **621,595 sheep** (-11.47%), while the number of holders declaring beef cattle went from 61,561 to 59,121 (-4.0%): more cattle, fewer hands. ```text PASS fewer holders declared beef cattle in 2025 (so 'fewer hands' holds): True vs True (exactly) — the two numbers above ``` ## Department by department ```python cat = q.query("kind == 'cattle'").groupby(["code", "year"])["animals"].sum().unstack() cat.columns = ["cattle_2024", "cattle_2025"] cat["change"] = cat["cattle_2025"] / cat["cattle_2024"] - 1 cat = cat.reset_index() cat["name"] = cat["code"].map(names) cat = cat.sort_values("change", ascending=False) c = cat.set_index("name")["change"] for name, expected in [("Salto", 0.059), ("Durazno", 0.058), ("Rivera", 0.052), ("Canelones", -0.048), ("Treinta y Tres", -0.046)]: cw.check(f"{name} cattle 2024 → 2025", round(c[name], 3), expected, tol=0.0005, source="expedition 2026-07-05") cc.show_table(cat[["name", "cattle_2024", "cattle_2025", "change"]], floatfmt="+.3f", name="cattle-change") ``` ```text PASS Salto cattle 2024 → 2025: 0.059 vs 0.059 (within ±0.0005) — expedition 2026-07-05 PASS Durazno cattle 2024 → 2025: 0.058 vs 0.058 (within ±0.0005) — expedition 2026-07-05 PASS Rivera cattle 2024 → 2025: 0.052 vs 0.052 (within ±0.0005) — expedition 2026-07-05 PASS Canelones cattle 2024 → 2025: -0.048 vs -0.048 (within ±0.0005) — expedition 2026-07-05 PASS Treinta y Tres cattle 2024 → 2025: -0.046 vs -0.046 (within ±0.0005) — expedition 2026-07-05 ``` | name | cattle_2024 | cattle_2025 | change | |:---------------|:--------------|:--------------|---------:| | Salto | 849,010 | 898,899 | 0.059 | | Durazno | 900,286 | 952,171 | 0.058 | | Rivera | 608,407 | 639,917 | 0.052 | | Flores | 393,191 | 409,429 | 0.041 | | Tacuarembó | 1,034,575 | 1,061,136 | 0.026 | | Florida | 1,095,226 | 1,117,097 | 0.02 | | Artigas | 730,165 | 742,198 | 0.016 | | Maldonado | 230,533 | 234,191 | 0.016 | | Lavalleja | 644,438 | 654,451 | 0.016 | | Montevideo | 2,699 | 2,736 | 0.014 | | Soriano | 570,050 | 567,452 | -0.005 | | Cerro Largo | 957,379 | 950,412 | -0.007 | | Rocha | 688,313 | 682,644 | -0.008 | | Paysandú | 809,798 | 801,512 | -0.01 | | Río Negro | 555,782 | 549,369 | -0.012 | | San José | 548,346 | 541,254 | -0.013 | | Colonia | 551,213 | 537,066 | -0.026 | | Treinta y Tres | 613,051 | 584,734 | -0.046 | | Canelones | 355,686 | 338,601 | -0.048 | ```python # The sentence below names regions: hold it to them. NORTH_CENTRE = {"Artigas", "Salto", "Rivera", "Tacuarembó", "Paysandú", "Durazno", "Cerro Largo", "Flores"} SOUTH_EAST = {"Canelones", "Montevideo", "Maldonado", "Rocha", "Treinta y Tres", "Lavalleja", "San José", "Colonia"} cw.check("the three fastest growers are in the north and centre", set(cat["name"].head(3)) <= NORTH_CENTRE, True, source=str(sorted(NORTH_CENTRE))) cw.check("the two steepest declines are in the south and east", set(cat["name"].tail(2)) <= SOUTH_EAST, True, source=str(sorted(SOUTH_EAST))) cw.finding("where", f"Growth came from the north and centre (**{cat.iloc[0]['name']}** {cat.iloc[0]['change']:+.1%}, " f"{cat.iloc[1]['name']} {cat.iloc[1]['change']:+.1%}, {cat.iloc[2]['name']} {cat.iloc[2]['change']:+.1%}); " f"the south and east shrank (**{cat.iloc[-1]['name']}** {cat.iloc[-1]['change']:+.1%}, " f"{cat.iloc[-2]['name']} {cat.iloc[-2]['change']:+.1%}).") ``` ```text PASS the three fastest growers are in the north and centre: True vs True (exactly) — ['Artigas', 'Cerro Largo', 'Durazno', 'Flores', 'Paysandú', 'Rivera', 'Salto', 'Tacuarembó'] PASS the two steepest declines are in the south and east: True vs True (exactly) — ['Canelones', 'Colonia', 'Lavalleja', 'Maldonado', 'Montevideo', 'Rocha', 'San José', 'Treinta y Tres'] ``` > **Finding.** Growth came from the north and centre (**Salto** +5.9%, Durazno +5.8%, Rivera +5.2%); the south and east shrank (**Canelones** -4.8%, Treinta y Tres -4.6%). ```python cc.show(cc.choropleth(cat, "change", title="Change in cattle, 2024 → 2025", subtitle="beef + dairy, PD_AD; blue is growth, brown is decline", legend="change", fmt="+.1%", diverging=True), "cattle-change-map", thumbnail=True, alt_text="Diverging map: the north and centre (Salto, Durazno, Rivera) gained the most cattle; Canelones and Treinta y Tres lost the most.") print(cc.map_credit()) ``` ![Diverging map: the north and centre (Salto, Durazno, Rivera) gained the most cattle; Canelones and Treinta y Tres lost the most.](charts/cattle-change-map.svg) Chart data and encoding: [charts/cattle-change-map.vl.json](charts/cattle-change-map.vl.json) (Vega-Lite). ```text Map: Natural Earth 1:10m admin-1 (public domain, v5.1.2), joined on the DICOSE department code. Made with Natural Earth. ``` ### The same change as a slope chart Each line is a department, from its 2024 to its 2025 herd (log scale, so a line's slope is its percentage change whatever the herd's size). ```python slope = cat.melt(id_vars=["name", "change"], value_vars=["cattle_2024", "cattle_2025"], var_name="year", value_name="cattle") slope["year"] = slope["year"].str[-4:] slope["direction"] = slope["change"].map(lambda v: "grew" if v > 0 else "shrank") base = alt.Chart(slope).encode( x=alt.X("year:N", title=None, axis=alt.Axis(labelAngle=0)), y=alt.Y("cattle:Q", scale=alt.Scale(type="log"), title="cattle (log scale)", axis=alt.Axis(format="~s")), color=alt.Color("direction:N", scale=alt.Scale(domain=["grew", "shrank"], range=[cc.CATTLE, cc.FALL]), title=None), detail="name:N", tooltip=["name:N", alt.Tooltip("cattle:Q", format=","), alt.Tooltip("change:Q", format="+.1%")], ) lines = base.mark_line(point=True) lab = alt.Chart(slope).transform_filter("datum.year == '2025'").mark_text(align="left", dx=6, fontSize=10).encode( x=alt.X("year:N"), y=alt.Y("cattle:Q", scale=alt.Scale(type="log")), text="name:N") cc.show((lines + lab).properties(width=260, height=440, title=alt.Title("Cattle by department, 2024 → 2025")), "cattle-slope", alt_text="Slope chart: most department lines rise gently; Canelones and Treinta y Tres fall.") ``` ![Slope chart: most department lines rise gently; Canelones and Treinta y Tres fall.](charts/cattle-slope.svg) Chart data and encoding: [charts/cattle-slope.vl.json](charts/cattle-slope.vl.json) (Vega-Lite). ## Where every number came from ```python cc.show_table(cw.checks()[["check", "got", "expected", "rule", "source"]], name="checks") cw.provenance() ``` | check | got | expected | rule | source | |:--------------------------------------------------------------------|:-------|:-----------|:---------------|:--------------------------------------------------------------------------------------------------------| | both years passed their quality gate | True | True | exactly | cow:QualityReport | | no blocking violations in either year | 0 | 0 | exactly | cow:QualityReport | | every counted warning is recorded | 4 | 4 | exactly | cow:warningCount vs cow:QualityWarning rows | | no warning touches the dimensions this comparison uses | [] | [] | exactly | uses: ['Datos animales', 'datos_animales', 'departments', 'species'] | | 2024 holders declaring beef cattle: cube = CSV | 61561 | 61561 | exactly | datosanimales.csv 2024, CantidadTenedoresPorEspecie | | 2025 holders declaring beef cattle: cube = CSV | 59121 | 59121 | exactly | datosanimales.csv 2025, CantidadTenedoresPorEspecie | | cattle 2024 → 2025 | 0.0105 | 0.0105 | within ±5e-05 | expedition 2026-07-05: +1.05% | | fewer holders declared beef cattle in 2025 (so 'fewer hands' holds) | True | True | exactly | the two numbers above | | Salto cattle 2024 → 2025 | 0.059 | 0.059 | within ±0.0005 | expedition 2026-07-05 | | Durazno cattle 2024 → 2025 | 0.058 | 0.058 | within ±0.0005 | expedition 2026-07-05 | | Rivera cattle 2024 → 2025 | 0.052 | 0.052 | within ±0.0005 | expedition 2026-07-05 | | Canelones cattle 2024 → 2025 | -0.048 | -0.048 | within ±0.0005 | expedition 2026-07-05 | | Treinta y Tres cattle 2024 → 2025 | -0.046 | -0.046 | within ±0.0005 | expedition 2026-07-05 | | the three fastest growers are in the north and centre | True | True | exactly | ['Artigas', 'Cerro Largo', 'Durazno', 'Flores', 'Paysandú', 'Rivera', 'Salto', 'Tacuarembó'] | | the two steepest declines are in the south and east | True | True | exactly | ['Canelones', 'Colonia', 'Lavalleja', 'Maldonado', 'Montevideo', 'Rocha', 'San José', 'Treinta y Tres'] | ```text kind target rows \ 0 sparql uruguay-cows 2 1 sparql uruguay-cows 4 2 sparql uruguay-cows 228 first line 0 SELECT ?year ?passed ?blocking ?warnings 1 SELECT ?year ?message 2 SELECT ?year ?department ?species (SUM(?n) AS ... ``` ```python cw.save() ```