fix: resolve type mismatch errors across WDL files by claymcleod · Pull Request #1816 · broadinstitute/warp

claymcleod · 2026-04-11T18:01:30Z

Fixes type mismatch errors flagged by sprocket lint across 5 files. These fall into a few categories:

Int-to-String coercion. range() returns Array[Int], but some task inputs expect String id. We use string interpolation ("~{i}") to convert.
Conditional optional unwrapping. The pattern if defined(x) then x else y doesn't narrow the type of x from T? to T in the then branch per the WDL spec. Replaced with select_first([x, y]).
Incorrect declared types. String java_memory_size = (memory_size - 1) * 1000 assigns an Int expression to a String — changed to Int. Similarly, Array[Int] mem_gb_per_chunk = read_lines(...) is invalid since read_lines always returns Array[String].
Boolean in Map[String, String]. InputQC.passes_qc is Boolean, but the map literal expects String values. Wrapped with "~{...}".
read_lines to Array[Int]. WDL has no String → Int coercion in any spec version (checked 1.0, 1.1, and 1.2). Changed GetFingerprintingIntervalIndices to write JSON and use read_json() so the output can be typed as Array[Int] for array indexing.

Note that there are remaining type mismatch errors related to passing optional types (T?) to non-optional call inputs that have defaults (T x = default). These are left as-is because the code is actually correct — passing None to a defaulted input means "use the default." This behavior was formalized in the WDL 1.2 spec (openwdl/wdl#462, openwdl/wdl#634) and is already implemented by Cromwell and miniwdl for all WDL versions. openwdl/wdl#761 tracks backporting this clarification to the 1.0 and 1.1 specs, after which Sprocket will implement it and these errors will resolve.

Depends on #1815.

Placeholders in WDL cannot contain more than one option (e.g., `default` + `sep` or `true`/`false` paired with a second `default` + `sep` placeholder). This rewrites those expressions using `prefix()`, `sep()`, and `select_first()` to satisfy `sprocket lint` while preserving the original command-line output.

Placeholder options like `default` expect string values. Changes `default=0` to `default="0"` and `default=250` to `default="250"` across two task files.

…n 1.0 Adds `version 1.0` declarations, wraps inputs in `input {}` blocks, and changes `cpu` runtime values from strings to bare `Int` variables across seven task/workflow files.

claymcleod

Inline comments explaining each fix for reviewer context.

claymcleod · 2026-04-11T18:03:00Z

                vcf_intervals_idx = training_vcf_so_bgz_idx,
                intersecting_intervals=intersecting_intervals,
-                id = i
+                id = "~{i}"


range() returns Array[Int], but the sitesOnlyAndHQFilterVcf task declares String id. WDL doesn't coerce Int to String implicitly, so we use string interpolation to convert.

claymcleod · 2026-04-11T18:03:00Z

                output_name="full_data_sites_filtered",
                service_account_json=service_account_json,
-                id = j
+                id = "~{j}"


Same Int to String fix as above — j comes from range() but filter_by_sites_only expects String id.

claymcleod · 2026-04-11T18:03:00Z

    String output_filename_idx = output_filename + ".tbi"
    String has_service_account_file = if (defined(service_account_json)) then 'true' else 'false'
-    String service_account_basename_pre = if (defined(service_account_json)) then service_account_json else ''
+    String service_account_basename_pre = select_first([service_account_json, ''])


The pattern if defined(x) then x else y doesn't narrow x from String? to String in the then branch per the WDL spec, so Sprocket flags the result as String? being assigned to String. select_first([x, y]) does the same thing and returns the correct non-optional type.

claymcleod · 2026-04-11T18:03:00Z

    String output_filename_idx = output_filename + ".tbi"
    String has_service_account_file = if (defined(service_account_json)) then 'true' else 'false'
-    String service_account_basename_pre = if (defined(service_account_json)) then service_account_json else ''
+    String service_account_basename_pre = select_first([service_account_json, ''])


Same select_first fix as in sitesOnlyAndHQFilterVcf above — identical pattern in the filter_by_sites_only task.

claymcleod · 2026-04-11T18:03:01Z

    output {
        Array[String] reference_chunk_file_paths = read_lines("reference_shard_file_paths.tsv")
-        Array[Int] mem_gb_per_chunk = read_lines("memory_per_chunk.tsv")
+        Array[String] mem_gb_per_chunk = read_lines("memory_per_chunk.tsv")


read_lines() always returns Array[String] per the WDL spec. This output isn't used anywhere in the workflow currently, so the type change is safe. If it's consumed downstream in the future, the values are integer strings from int(np.ceil(x)) in the Python command.

claymcleod · 2026-04-11T18:03:01Z

  # Compute memory to use based on the CPU count, following the pattern of
  # 3.75GiB / cpu used by GCP's pricing: https://cloud.google.com/compute/pricing
-  Int memory = if defined(machine_mem_mb) then machine_mem_mb else round(cpu * 3.75 * 1024)
+  Int memory = select_first([machine_mem_mb, round(cpu * 3.75 * 1024)])


machine_mem_mb is Int? (line 886). The if defined(x) then x else y pattern doesn't narrow the optional, so the result type is Int? rather than Int. select_first does the same thing with correct typing.

claymcleod · 2026-04-11T18:03:01Z

+
+    python3 -c "
+with open('indices.out') as f:
+    indices = [int(line.strip()) for line in f if line.strip()]


The original output used read_lines() which returns Array[String], but callers (JointGenotyping.wdl:397, UltimaGenomicsJointGenotyping.wdl:269) need Array[Int] for array indexing. WDL has no String → Int coercion in any spec version, so we convert the indices file to JSON and use read_json() which can deserialize directly to Array[Int].

This is a little more involved—I have not tested it myself, but perhaps someone should.

claymcleod · 2026-04-11T18:03:01Z


  Int memory_size = ceil((if (disk_size < 110) then 5 else 7) * memory_multiplier)
-  String java_memory_size = (memory_size - 1) * 1000
+  Int java_memory_size = (memory_size - 1) * 1000


The RHS (memory_size - 1) * 1000 is Int arithmetic, but the declaration was String. This worked at runtime because the value gets interpolated into the command string anyway, but the declared type should match the expression type.

claymcleod · 2026-04-11T18:03:01Z

      input:
        input_map = {
-          "passes_qc": InputQC.passes_qc,
+          "passes_qc": "~{InputQC.passes_qc}",


InputQC.passes_qc is Boolean, but this is a Map[String, String] literal so all values must be String. Wrapping in "~{...}" converts the boolean to its string representation ("true" or "false").

Fixes 16 type mismatch lint errors found by sprocket across 10 files. The errors fall into several categories: - Optional-to-non-optional mismatches: uses `select_first()` to unwrap optional values when passing to task inputs that have defaults - Int-to-String coercion: uses string interpolation (`"~{i}"`) to convert `Int` values from `range()` to `String` task inputs - Conditional optional unwrapping: replaces `if defined(x) then x else y` with `select_first([x, y])` to avoid type narrowing issues - Incorrect declared types: changes `String` to `Int` where the RHS is arithmetic, and `Array[Int]` to `Array[String]` for `read_lines()` - Boolean-in-Map coercion: uses string interpolation to convert a `Boolean` value to `String` for a `Map[String, String]` literal - read_lines-to-Array[Int]: converts task output to use `read_json()` since WDL has no `String` to `Int` coercion in any spec version

jessicaway · 2026-05-06T14:51:08Z

@claymcleod Thanks so much for this contribution and highlighting the syntax issue here. Unfortunately, we do not yet have tests for any of the AoU pipelines and cannot change them at this time. We will keep this open to revisit in the future.

claymcleod · 2026-05-07T18:10:53Z

Thanks @jessicaway. It would be great if these workflows could be used more generally, please let me know if there's anything I can do to help.

rsc3 · 2026-06-02T14:22:30Z

Hello,

If you could remove (or start a new PR) any of these changes in the AoU directory here:

all_of_us/**

then we can go ahead and merge the changes across all of the (non-beta) pipelines outside this directory.

Thank you!

claymcleod added 4 commits April 11, 2026 01:08

fix: uses Int instead of String for cpu runtime values

31a8599

fix: quote integer literals in placeholder default options

661f7f8

Placeholder options like `default` expect string values. Changes `default=0` to `default="0"` and `default=250` to `default="250"` across two task files.

fix: convert draft-2 WDL files in all_of_us/rna_seq/GTEx/ to versio…

1956b7f

…n 1.0 Adds `version 1.0` declarations, wraps inputs in `input {}` blocks, and changes `cpu` runtime values from strings to bare `Int` variables across seven task/workflow files.

claymcleod commented Apr 11, 2026

View reviewed changes

claymcleod force-pushed the fix/type-mismatches branch from 492db6f to db8e352 Compare April 11, 2026 19:23

claymcleod marked this pull request as ready for review April 11, 2026 19:36

claymcleod mentioned this pull request Apr 11, 2026

fix: resolves naming conflicts, expression type errors, and misc lint issues #1817

Open

Uh oh!

Conversation

claymcleod commented Apr 11, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

claymcleod left a comment

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

claymcleod Apr 11, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

claymcleod Apr 11, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

jessicaway commented May 6, 2026

Uh oh!

claymcleod commented May 7, 2026

Uh oh!

rsc3 commented Jun 2, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

claymcleod commented Apr 11, 2026 •

edited

Loading

claymcleod Apr 11, 2026 •

edited

Loading

claymcleod Apr 11, 2026 •

edited

Loading